diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,15607 @@
+{
+  "best_metric": 0.07300831,
+  "best_model_checkpoint": "/home/patrickbarker/output/qwen2-vl-7b-instruct/v0-20250101-185154/checkpoint-6036",
+  "epoch": 14.0,
+  "eval_steps": 200,
+  "global_step": 7759,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "acc": 0.66110599,
+      "epoch": 0.0003313452617627568,
+      "grad_norm": 50.27714337190847,
+      "learning_rate": 0.0,
+      "loss": 3.7548461,
+      "memory(GiB)": 31.98,
+      "step": 1,
+      "train_speed(iter/s)": 0.051634
+    },
+    {
+      "acc": 0.65853631,
+      "epoch": 0.0016567263088137839,
+      "grad_norm": 59.248621100065584,
+      "learning_rate": 2.0086965557947446e-06,
+      "loss": 3.62318873,
+      "memory(GiB)": 33.01,
+      "step": 5,
+      "train_speed(iter/s)": 0.124439
+    },
+    {
+      "acc": 0.7222682,
+      "epoch": 0.0033134526176275677,
+      "grad_norm": 21.791730292901168,
+      "learning_rate": 2.8737950746583047e-06,
+      "loss": 1.76198959,
+      "memory(GiB)": 33.01,
+      "step": 10,
+      "train_speed(iter/s)": 0.15095
+    },
+    {
+      "acc": 0.77555671,
+      "epoch": 0.004970178926441352,
+      "grad_norm": 18.163931905418693,
+      "learning_rate": 3.379845267622901e-06,
+      "loss": 1.16682892,
+      "memory(GiB)": 33.01,
+      "step": 15,
+      "train_speed(iter/s)": 0.162546
+    },
+    {
+      "acc": 0.79614887,
+      "epoch": 0.0066269052352551355,
+      "grad_norm": 23.170639029991356,
+      "learning_rate": 3.738893593521864e-06,
+      "loss": 1.05961266,
+      "memory(GiB)": 33.01,
+      "step": 20,
+      "train_speed(iter/s)": 0.16897
+    },
+    {
+      "acc": 0.79792585,
+      "epoch": 0.00828363154406892,
+      "grad_norm": 25.79687222394801,
+      "learning_rate": 4.017393111589489e-06,
+      "loss": 1.11055069,
+      "memory(GiB)": 33.01,
+      "step": 25,
+      "train_speed(iter/s)": 0.173184
+    },
+    {
+      "acc": 0.80801973,
+      "epoch": 0.009940357852882704,
+      "grad_norm": 15.051157865550348,
+      "learning_rate": 4.24494378648646e-06,
+      "loss": 0.99607706,
+      "memory(GiB)": 33.01,
+      "step": 30,
+      "train_speed(iter/s)": 0.176052
+    },
+    {
+      "acc": 0.82084551,
+      "epoch": 0.011597084161696488,
+      "grad_norm": 10.223050479974178,
+      "learning_rate": 4.437335140791102e-06,
+      "loss": 0.95394745,
+      "memory(GiB)": 33.01,
+      "step": 35,
+      "train_speed(iter/s)": 0.17821
+    },
+    {
+      "acc": 0.83056355,
+      "epoch": 0.013253810470510271,
+      "grad_norm": 11.864398422104427,
+      "learning_rate": 4.603992112385424e-06,
+      "loss": 0.87341337,
+      "memory(GiB)": 33.01,
+      "step": 40,
+      "train_speed(iter/s)": 0.179859
+    },
+    {
+      "acc": 0.81267252,
+      "epoch": 0.014910536779324055,
+      "grad_norm": 12.40283491772095,
+      "learning_rate": 4.750993979451057e-06,
+      "loss": 0.95986633,
+      "memory(GiB)": 33.01,
+      "step": 45,
+      "train_speed(iter/s)": 0.181182
+    },
+    {
+      "acc": 0.83013372,
+      "epoch": 0.01656726308813784,
+      "grad_norm": 10.434349025985295,
+      "learning_rate": 4.882491630453049e-06,
+      "loss": 0.87351093,
+      "memory(GiB)": 33.01,
+      "step": 50,
+      "train_speed(iter/s)": 0.182219
+    },
+    {
+      "acc": 0.83825665,
+      "epoch": 0.018223989396951624,
+      "grad_norm": 9.483246391000499,
+      "learning_rate": 5.001445725187638e-06,
+      "loss": 0.84431524,
+      "memory(GiB)": 33.01,
+      "step": 55,
+      "train_speed(iter/s)": 0.183104
+    },
+    {
+      "acc": 0.81640253,
+      "epoch": 0.019880715705765408,
+      "grad_norm": 10.776285793721472,
+      "learning_rate": 5.11004230535002e-06,
+      "loss": 0.95539846,
+      "memory(GiB)": 33.01,
+      "step": 60,
+      "train_speed(iter/s)": 0.183885
+    },
+    {
+      "acc": 0.82131834,
+      "epoch": 0.021537442014579192,
+      "grad_norm": 8.627366353102206,
+      "learning_rate": 5.209941475102492e-06,
+      "loss": 0.92218132,
+      "memory(GiB)": 33.01,
+      "step": 65,
+      "train_speed(iter/s)": 0.184523
+    },
+    {
+      "acc": 0.81877251,
+      "epoch": 0.023194168323392977,
+      "grad_norm": 7.73227486810335,
+      "learning_rate": 5.302433659654663e-06,
+      "loss": 0.91788635,
+      "memory(GiB)": 33.01,
+      "step": 70,
+      "train_speed(iter/s)": 0.185058
+    },
+    {
+      "acc": 0.83363972,
+      "epoch": 0.02485089463220676,
+      "grad_norm": 10.385061722217497,
+      "learning_rate": 5.388541823417645e-06,
+      "loss": 0.89225368,
+      "memory(GiB)": 33.01,
+      "step": 75,
+      "train_speed(iter/s)": 0.185511
+    },
+    {
+      "acc": 0.8337285,
+      "epoch": 0.026507620941020542,
+      "grad_norm": 7.860817957230683,
+      "learning_rate": 5.469090631248983e-06,
+      "loss": 0.87068348,
+      "memory(GiB)": 33.01,
+      "step": 80,
+      "train_speed(iter/s)": 0.185937
+    },
+    {
+      "acc": 0.84761944,
+      "epoch": 0.028164347249834326,
+      "grad_norm": 9.115430859362677,
+      "learning_rate": 5.5447546056702515e-06,
+      "loss": 0.81180515,
+      "memory(GiB)": 33.01,
+      "step": 85,
+      "train_speed(iter/s)": 0.186316
+    },
+    {
+      "acc": 0.84264908,
+      "epoch": 0.02982107355864811,
+      "grad_norm": 13.663567397042407,
+      "learning_rate": 5.616092498314617e-06,
+      "loss": 0.88545876,
+      "memory(GiB)": 33.01,
+      "step": 90,
+      "train_speed(iter/s)": 0.186635
+    },
+    {
+      "acc": 0.84530535,
+      "epoch": 0.031477799867461895,
+      "grad_norm": 7.4754005630581135,
+      "learning_rate": 5.683572355914555e-06,
+      "loss": 0.82264423,
+      "memory(GiB)": 33.01,
+      "step": 95,
+      "train_speed(iter/s)": 0.186926
+    },
+    {
+      "acc": 0.84374695,
+      "epoch": 0.03313452617627568,
+      "grad_norm": 11.99236287767985,
+      "learning_rate": 5.7475901493166095e-06,
+      "loss": 0.78904347,
+      "memory(GiB)": 33.01,
+      "step": 100,
+      "train_speed(iter/s)": 0.187211
+    },
+    {
+      "acc": 0.84980297,
+      "epoch": 0.03479125248508946,
+      "grad_norm": 7.905497376563473,
+      "learning_rate": 5.808483852619259e-06,
+      "loss": 0.77175064,
+      "memory(GiB)": 33.01,
+      "step": 105,
+      "train_speed(iter/s)": 0.187455
+    },
+    {
+      "acc": 0.84676991,
+      "epoch": 0.03644797879390325,
+      "grad_norm": 8.403941397142672,
+      "learning_rate": 5.8665442440511985e-06,
+      "loss": 0.7824954,
+      "memory(GiB)": 33.01,
+      "step": 110,
+      "train_speed(iter/s)": 0.187681
+    },
+    {
+      "acc": 0.83980999,
+      "epoch": 0.03810470510271703,
+      "grad_norm": 8.971608842298808,
+      "learning_rate": 5.922023303967215e-06,
+      "loss": 0.83844948,
+      "memory(GiB)": 33.01,
+      "step": 115,
+      "train_speed(iter/s)": 0.187886
+    },
+    {
+      "acc": 0.83745537,
+      "epoch": 0.039761431411530816,
+      "grad_norm": 11.092540697396046,
+      "learning_rate": 5.97514082421358e-06,
+      "loss": 0.84876623,
+      "memory(GiB)": 33.01,
+      "step": 120,
+      "train_speed(iter/s)": 0.18808
+    },
+    {
+      "acc": 0.8322444,
+      "epoch": 0.0414181577203446,
+      "grad_norm": 3.396144727915799,
+      "learning_rate": 6.026089667384234e-06,
+      "loss": 0.85240536,
+      "memory(GiB)": 33.01,
+      "step": 125,
+      "train_speed(iter/s)": 0.18825
+    },
+    {
+      "acc": 0.83602228,
+      "epoch": 0.043074884029158385,
+      "grad_norm": 8.270916567528229,
+      "learning_rate": 6.075039993966052e-06,
+      "loss": 0.82148342,
+      "memory(GiB)": 33.01,
+      "step": 130,
+      "train_speed(iter/s)": 0.188408
+    },
+    {
+      "acc": 0.81821384,
+      "epoch": 0.04473161033797217,
+      "grad_norm": 7.753889762215579,
+      "learning_rate": 6.122142691279214e-06,
+      "loss": 0.88220882,
+      "memory(GiB)": 33.01,
+      "step": 135,
+      "train_speed(iter/s)": 0.188567
+    },
+    {
+      "acc": 0.84211292,
+      "epoch": 0.04638833664678595,
+      "grad_norm": 7.027352256083138,
+      "learning_rate": 6.167532178518222e-06,
+      "loss": 0.83544054,
+      "memory(GiB)": 33.01,
+      "step": 140,
+      "train_speed(iter/s)": 0.188713
+    },
+    {
+      "acc": 0.83703737,
+      "epoch": 0.04804506295559974,
+      "grad_norm": 6.462244441845264,
+      "learning_rate": 6.211328719346929e-06,
+      "loss": 0.80138245,
+      "memory(GiB)": 33.01,
+      "step": 145,
+      "train_speed(iter/s)": 0.188826
+    },
+    {
+      "acc": 0.85498161,
+      "epoch": 0.04970178926441352,
+      "grad_norm": 8.758579967575177,
+      "learning_rate": 6.253640342281205e-06,
+      "loss": 0.77901378,
+      "memory(GiB)": 33.01,
+      "step": 150,
+      "train_speed(iter/s)": 0.188819
+    },
+    {
+      "acc": 0.85058413,
+      "epoch": 0.051358515573227306,
+      "grad_norm": 17.338366738017733,
+      "learning_rate": 6.2945644460697445e-06,
+      "loss": 0.7858984,
+      "memory(GiB)": 33.01,
+      "step": 155,
+      "train_speed(iter/s)": 0.188946
+    },
+    {
+      "acc": 0.84381838,
+      "epoch": 0.053015241882041084,
+      "grad_norm": 5.435548438025349,
+      "learning_rate": 6.3341891501125445e-06,
+      "loss": 0.77058525,
+      "memory(GiB)": 33.01,
+      "step": 160,
+      "train_speed(iter/s)": 0.18896
+    },
+    {
+      "acc": 0.83954039,
+      "epoch": 0.05467196819085487,
+      "grad_norm": 4.399997862609139,
+      "learning_rate": 6.372594437015794e-06,
+      "loss": 0.80552349,
+      "memory(GiB)": 33.01,
+      "step": 165,
+      "train_speed(iter/s)": 0.189058
+    },
+    {
+      "acc": 0.85681477,
+      "epoch": 0.05632869449966865,
+      "grad_norm": 5.206696585986359,
+      "learning_rate": 6.409853124533811e-06,
+      "loss": 0.73895197,
+      "memory(GiB)": 33.01,
+      "step": 170,
+      "train_speed(iter/s)": 0.189155
+    },
+    {
+      "acc": 0.85199137,
+      "epoch": 0.05798542080848244,
+      "grad_norm": 4.4333768843681725,
+      "learning_rate": 6.446031696585847e-06,
+      "loss": 0.77676201,
+      "memory(GiB)": 33.01,
+      "step": 175,
+      "train_speed(iter/s)": 0.189245
+    },
+    {
+      "acc": 0.85372562,
+      "epoch": 0.05964214711729622,
+      "grad_norm": 17.001757223516282,
+      "learning_rate": 6.4811910171781766e-06,
+      "loss": 0.76763196,
+      "memory(GiB)": 33.01,
+      "step": 180,
+      "train_speed(iter/s)": 0.189335
+    },
+    {
+      "acc": 0.84696083,
+      "epoch": 0.061298873426110005,
+      "grad_norm": 7.799436298527937,
+      "learning_rate": 6.515386946489136e-06,
+      "loss": 0.76351337,
+      "memory(GiB)": 33.01,
+      "step": 185,
+      "train_speed(iter/s)": 0.189422
+    },
+    {
+      "acc": 0.85822506,
+      "epoch": 0.06295559973492379,
+      "grad_norm": 11.328352996137795,
+      "learning_rate": 6.548670874778115e-06,
+      "loss": 0.79490757,
+      "memory(GiB)": 33.01,
+      "step": 190,
+      "train_speed(iter/s)": 0.189494
+    },
+    {
+      "acc": 0.85514393,
+      "epoch": 0.06461232604373758,
+      "grad_norm": 6.912811704629352,
+      "learning_rate": 6.581090186930649e-06,
+      "loss": 0.7586585,
+      "memory(GiB)": 33.01,
+      "step": 195,
+      "train_speed(iter/s)": 0.189567
+    },
+    {
+      "acc": 0.84247036,
+      "epoch": 0.06626905235255136,
+      "grad_norm": 5.22561576153318,
+      "learning_rate": 6.612688668180168e-06,
+      "loss": 0.811413,
+      "memory(GiB)": 33.01,
+      "step": 200,
+      "train_speed(iter/s)": 0.189625
+    },
+    {
+      "acc": 0.84955778,
+      "epoch": 0.06792577866136515,
+      "grad_norm": 7.588959185273035,
+      "learning_rate": 6.643506859724346e-06,
+      "loss": 0.76285973,
+      "memory(GiB)": 33.01,
+      "step": 205,
+      "train_speed(iter/s)": 0.189696
+    },
+    {
+      "acc": 0.84231005,
+      "epoch": 0.06958250497017893,
+      "grad_norm": 5.213604633751543,
+      "learning_rate": 6.6735823714828186e-06,
+      "loss": 0.76322999,
+      "memory(GiB)": 33.01,
+      "step": 210,
+      "train_speed(iter/s)": 0.189766
+    },
+    {
+      "acc": 0.8555872,
+      "epoch": 0.0712392312789927,
+      "grad_norm": 5.17769941057491,
+      "learning_rate": 6.702950158049068e-06,
+      "loss": 0.73766384,
+      "memory(GiB)": 33.01,
+      "step": 215,
+      "train_speed(iter/s)": 0.189824
+    },
+    {
+      "acc": 0.85509701,
+      "epoch": 0.0728959575878065,
+      "grad_norm": 9.338793341973062,
+      "learning_rate": 6.731642762914758e-06,
+      "loss": 0.74758358,
+      "memory(GiB)": 33.01,
+      "step": 220,
+      "train_speed(iter/s)": 0.18988
+    },
+    {
+      "acc": 0.85551167,
+      "epoch": 0.07455268389662027,
+      "grad_norm": 7.799786407663588,
+      "learning_rate": 6.759690535245802e-06,
+      "loss": 0.72227631,
+      "memory(GiB)": 33.01,
+      "step": 225,
+      "train_speed(iter/s)": 0.189939
+    },
+    {
+      "acc": 0.85811892,
+      "epoch": 0.07620941020543406,
+      "grad_norm": 5.997515776481016,
+      "learning_rate": 6.7871218228307745e-06,
+      "loss": 0.72223263,
+      "memory(GiB)": 33.01,
+      "step": 230,
+      "train_speed(iter/s)": 0.18999
+    },
+    {
+      "acc": 0.85784416,
+      "epoch": 0.07786613651424784,
+      "grad_norm": 8.41060544225861,
+      "learning_rate": 6.813963144277111e-06,
+      "loss": 0.77040291,
+      "memory(GiB)": 33.01,
+      "step": 235,
+      "train_speed(iter/s)": 0.190035
+    },
+    {
+      "acc": 0.85461082,
+      "epoch": 0.07952286282306163,
+      "grad_norm": 8.220098968672605,
+      "learning_rate": 6.84023934307714e-06,
+      "loss": 0.75113549,
+      "memory(GiB)": 33.01,
+      "step": 240,
+      "train_speed(iter/s)": 0.190082
+    },
+    {
+      "acc": 0.84146032,
+      "epoch": 0.08117958913187541,
+      "grad_norm": 5.976461123958971,
+      "learning_rate": 6.865973725787461e-06,
+      "loss": 0.78689213,
+      "memory(GiB)": 33.01,
+      "step": 245,
+      "train_speed(iter/s)": 0.190128
+    },
+    {
+      "acc": 0.85554123,
+      "epoch": 0.0828363154406892,
+      "grad_norm": 3.9102866660697404,
+      "learning_rate": 6.891188186247792e-06,
+      "loss": 0.73804913,
+      "memory(GiB)": 33.01,
+      "step": 250,
+      "train_speed(iter/s)": 0.190174
+    },
+    {
+      "acc": 0.85871716,
+      "epoch": 0.08449304174950298,
+      "grad_norm": 10.08654606800113,
+      "learning_rate": 6.915903317498407e-06,
+      "loss": 0.72549114,
+      "memory(GiB)": 33.01,
+      "step": 255,
+      "train_speed(iter/s)": 0.190215
+    },
+    {
+      "acc": 0.83968134,
+      "epoch": 0.08614976805831677,
+      "grad_norm": 7.7105875410612565,
+      "learning_rate": 6.940138512829612e-06,
+      "loss": 0.84207258,
+      "memory(GiB)": 33.01,
+      "step": 260,
+      "train_speed(iter/s)": 0.190248
+    },
+    {
+      "acc": 0.85006838,
+      "epoch": 0.08780649436713055,
+      "grad_norm": 5.088348470726073,
+      "learning_rate": 6.963912057205657e-06,
+      "loss": 0.78956671,
+      "memory(GiB)": 33.01,
+      "step": 265,
+      "train_speed(iter/s)": 0.190286
+    },
+    {
+      "acc": 0.8407608,
+      "epoch": 0.08946322067594434,
+      "grad_norm": 7.157954060997024,
+      "learning_rate": 6.987241210142773e-06,
+      "loss": 0.80997934,
+      "memory(GiB)": 33.01,
+      "step": 270,
+      "train_speed(iter/s)": 0.190318
+    },
+    {
+      "acc": 0.84329042,
+      "epoch": 0.09111994698475812,
+      "grad_norm": 4.080799357922467,
+      "learning_rate": 7.010142280982383e-06,
+      "loss": 0.78907938,
+      "memory(GiB)": 33.01,
+      "step": 275,
+      "train_speed(iter/s)": 0.19035
+    },
+    {
+      "acc": 0.8445241,
+      "epoch": 0.0927766732935719,
+      "grad_norm": 4.975765769175333,
+      "learning_rate": 7.0326306973817816e-06,
+      "loss": 0.79310322,
+      "memory(GiB)": 33.01,
+      "step": 280,
+      "train_speed(iter/s)": 0.190378
+    },
+    {
+      "acc": 0.82853336,
+      "epoch": 0.09443339960238568,
+      "grad_norm": 3.8534560662201405,
+      "learning_rate": 7.054721067742711e-06,
+      "loss": 0.82995491,
+      "memory(GiB)": 33.01,
+      "step": 285,
+      "train_speed(iter/s)": 0.19042
+    },
+    {
+      "acc": 0.84708948,
+      "epoch": 0.09609012591119948,
+      "grad_norm": 7.862309897547912,
+      "learning_rate": 7.076427238210489e-06,
+      "loss": 0.78673692,
+      "memory(GiB)": 33.01,
+      "step": 290,
+      "train_speed(iter/s)": 0.190458
+    },
+    {
+      "acc": 0.85356722,
+      "epoch": 0.09774685222001325,
+      "grad_norm": 7.267843653288115,
+      "learning_rate": 7.097762344800689e-06,
+      "loss": 0.74209976,
+      "memory(GiB)": 33.01,
+      "step": 295,
+      "train_speed(iter/s)": 0.190496
+    },
+    {
+      "acc": 0.85457726,
+      "epoch": 0.09940357852882704,
+      "grad_norm": 4.996312399649092,
+      "learning_rate": 7.118738861144765e-06,
+      "loss": 0.73887339,
+      "memory(GiB)": 33.01,
+      "step": 300,
+      "train_speed(iter/s)": 0.190532
+    },
+    {
+      "acc": 0.85430183,
+      "epoch": 0.10106030483764082,
+      "grad_norm": 9.04522225013425,
+      "learning_rate": 7.13936864228921e-06,
+      "loss": 0.73701906,
+      "memory(GiB)": 33.01,
+      "step": 305,
+      "train_speed(iter/s)": 0.190558
+    },
+    {
+      "acc": 0.84619389,
+      "epoch": 0.10271703114645461,
+      "grad_norm": 3.602019206226853,
+      "learning_rate": 7.159662964933303e-06,
+      "loss": 0.76839342,
+      "memory(GiB)": 33.01,
+      "step": 310,
+      "train_speed(iter/s)": 0.190594
+    },
+    {
+      "acc": 0.85931883,
+      "epoch": 0.10437375745526839,
+      "grad_norm": 6.644233975861472,
+      "learning_rate": 7.179632564447415e-06,
+      "loss": 0.72944298,
+      "memory(GiB)": 33.01,
+      "step": 315,
+      "train_speed(iter/s)": 0.190628
+    },
+    {
+      "acc": 0.85696793,
+      "epoch": 0.10603048376408217,
+      "grad_norm": 6.89152177234198,
+      "learning_rate": 7.199287668976103e-06,
+      "loss": 0.71098285,
+      "memory(GiB)": 33.01,
+      "step": 320,
+      "train_speed(iter/s)": 0.19066
+    },
+    {
+      "acc": 0.84302912,
+      "epoch": 0.10768721007289596,
+      "grad_norm": 13.225574415600349,
+      "learning_rate": 7.218638030897237e-06,
+      "loss": 0.8497488,
+      "memory(GiB)": 33.01,
+      "step": 325,
+      "train_speed(iter/s)": 0.190693
+    },
+    {
+      "acc": 0.85180035,
+      "epoch": 0.10934393638170974,
+      "grad_norm": 9.01186226545562,
+      "learning_rate": 7.237692955879354e-06,
+      "loss": 0.78130593,
+      "memory(GiB)": 33.01,
+      "step": 330,
+      "train_speed(iter/s)": 0.190723
+    },
+    {
+      "acc": 0.84967728,
+      "epoch": 0.11100066269052353,
+      "grad_norm": 4.417536183486161,
+      "learning_rate": 7.256461329754014e-06,
+      "loss": 0.72536817,
+      "memory(GiB)": 33.01,
+      "step": 335,
+      "train_speed(iter/s)": 0.190753
+    },
+    {
+      "acc": 0.87272768,
+      "epoch": 0.1126573889993373,
+      "grad_norm": 5.571498245437071,
+      "learning_rate": 7.274951643397372e-06,
+      "loss": 0.68704929,
+      "memory(GiB)": 33.01,
+      "step": 340,
+      "train_speed(iter/s)": 0.190773
+    },
+    {
+      "acc": 0.86904411,
+      "epoch": 0.1143141153081511,
+      "grad_norm": 5.480038545049546,
+      "learning_rate": 7.293172015795371e-06,
+      "loss": 0.64733562,
+      "memory(GiB)": 33.01,
+      "step": 345,
+      "train_speed(iter/s)": 0.1908
+    },
+    {
+      "acc": 0.88172531,
+      "epoch": 0.11597084161696487,
+      "grad_norm": 4.423695596891556,
+      "learning_rate": 7.311130215449407e-06,
+      "loss": 0.63284264,
+      "memory(GiB)": 33.01,
+      "step": 350,
+      "train_speed(iter/s)": 0.190817
+    },
+    {
+      "acc": 0.86758823,
+      "epoch": 0.11762756792577866,
+      "grad_norm": 11.959440699921517,
+      "learning_rate": 7.328833680263689e-06,
+      "loss": 0.71492028,
+      "memory(GiB)": 33.01,
+      "step": 355,
+      "train_speed(iter/s)": 0.190848
+    },
+    {
+      "acc": 0.85512447,
+      "epoch": 0.11928429423459244,
+      "grad_norm": 11.531180840267398,
+      "learning_rate": 7.346289536041736e-06,
+      "loss": 0.74734936,
+      "memory(GiB)": 33.01,
+      "step": 360,
+      "train_speed(iter/s)": 0.190867
+    },
+    {
+      "acc": 0.86098652,
+      "epoch": 0.12094102054340623,
+      "grad_norm": 6.38954821051966,
+      "learning_rate": 7.363504613707136e-06,
+      "loss": 0.6976438,
+      "memory(GiB)": 33.01,
+      "step": 365,
+      "train_speed(iter/s)": 0.190887
+    },
+    {
+      "acc": 0.85538502,
+      "epoch": 0.12259774685222001,
+      "grad_norm": 7.489897904281903,
+      "learning_rate": 7.380485465352696e-06,
+      "loss": 0.72213078,
+      "memory(GiB)": 33.01,
+      "step": 370,
+      "train_speed(iter/s)": 0.19091
+    },
+    {
+      "acc": 0.85435352,
+      "epoch": 0.1242544731610338,
+      "grad_norm": 6.5745061795079955,
+      "learning_rate": 7.397238379212389e-06,
+      "loss": 0.7297389,
+      "memory(GiB)": 33.01,
+      "step": 375,
+      "train_speed(iter/s)": 0.190928
+    },
+    {
+      "acc": 0.84575977,
+      "epoch": 0.12591119946984758,
+      "grad_norm": 6.3145621197054895,
+      "learning_rate": 7.413769393641675e-06,
+      "loss": 0.76131997,
+      "memory(GiB)": 33.01,
+      "step": 380,
+      "train_speed(iter/s)": 0.190915
+    },
+    {
+      "acc": 0.85743666,
+      "epoch": 0.12756792577866136,
+      "grad_norm": 5.939012809105342,
+      "learning_rate": 7.430084310183997e-06,
+      "loss": 0.71653161,
+      "memory(GiB)": 33.01,
+      "step": 385,
+      "train_speed(iter/s)": 0.190927
+    },
+    {
+      "acc": 0.84744682,
+      "epoch": 0.12922465208747516,
+      "grad_norm": 7.54414322312517,
+      "learning_rate": 7.4461887057942085e-06,
+      "loss": 0.7426198,
+      "memory(GiB)": 33.01,
+      "step": 390,
+      "train_speed(iter/s)": 0.19092
+    },
+    {
+      "acc": 0.84886322,
+      "epoch": 0.13088137839628894,
+      "grad_norm": 8.937659297884245,
+      "learning_rate": 7.46208794428338e-06,
+      "loss": 0.75673151,
+      "memory(GiB)": 33.01,
+      "step": 395,
+      "train_speed(iter/s)": 0.190942
+    },
+    {
+      "acc": 0.84283085,
+      "epoch": 0.13253810470510272,
+      "grad_norm": 7.224979630512225,
+      "learning_rate": 7.477787187043728e-06,
+      "loss": 0.76008615,
+      "memory(GiB)": 33.01,
+      "step": 400,
+      "train_speed(iter/s)": 0.190961
+    },
+    {
+      "acc": 0.85820484,
+      "epoch": 0.1341948310139165,
+      "grad_norm": 6.728197096262106,
+      "learning_rate": 7.493291403107369e-06,
+      "loss": 0.73872232,
+      "memory(GiB)": 33.01,
+      "step": 405,
+      "train_speed(iter/s)": 0.190981
+    },
+    {
+      "acc": 0.83822203,
+      "epoch": 0.1358515573227303,
+      "grad_norm": 11.090670435976039,
+      "learning_rate": 7.5086053785879055e-06,
+      "loss": 0.83083191,
+      "memory(GiB)": 33.01,
+      "step": 410,
+      "train_speed(iter/s)": 0.190997
+    },
+    {
+      "acc": 0.83120193,
+      "epoch": 0.13750828363154408,
+      "grad_norm": 7.014092801931885,
+      "learning_rate": 7.52373372554976e-06,
+      "loss": 0.81330185,
+      "memory(GiB)": 33.01,
+      "step": 415,
+      "train_speed(iter/s)": 0.191013
+    },
+    {
+      "acc": 0.8598774,
+      "epoch": 0.13916500994035785,
+      "grad_norm": 5.396661691435741,
+      "learning_rate": 7.538680890346378e-06,
+      "loss": 0.73769159,
+      "memory(GiB)": 33.01,
+      "step": 420,
+      "train_speed(iter/s)": 0.191023
+    },
+    {
+      "acc": 0.85436583,
+      "epoch": 0.14082173624917163,
+      "grad_norm": 5.4805303609602625,
+      "learning_rate": 7.5534511614649965e-06,
+      "loss": 0.74521694,
+      "memory(GiB)": 33.01,
+      "step": 425,
+      "train_speed(iter/s)": 0.191036
+    },
+    {
+      "acc": 0.87025623,
+      "epoch": 0.1424784625579854,
+      "grad_norm": 5.209954254433347,
+      "learning_rate": 7.568048676912627e-06,
+      "loss": 0.69971952,
+      "memory(GiB)": 33.01,
+      "step": 430,
+      "train_speed(iter/s)": 0.191054
+    },
+    {
+      "acc": 0.84417782,
+      "epoch": 0.1441351888667992,
+      "grad_norm": 5.511804934678127,
+      "learning_rate": 7.582477431175086e-06,
+      "loss": 0.77695208,
+      "memory(GiB)": 33.01,
+      "step": 435,
+      "train_speed(iter/s)": 0.191071
+    },
+    {
+      "acc": 0.85898895,
+      "epoch": 0.145791915175613,
+      "grad_norm": 7.816283737518623,
+      "learning_rate": 7.596741281778317e-06,
+      "loss": 0.6926734,
+      "memory(GiB)": 33.01,
+      "step": 440,
+      "train_speed(iter/s)": 0.191083
+    },
+    {
+      "acc": 0.84782066,
+      "epoch": 0.14744864148442677,
+      "grad_norm": 6.60501392294181,
+      "learning_rate": 7.610843955479014e-06,
+      "loss": 0.78016205,
+      "memory(GiB)": 33.01,
+      "step": 445,
+      "train_speed(iter/s)": 0.191102
+    },
+    {
+      "acc": 0.85637875,
+      "epoch": 0.14910536779324055,
+      "grad_norm": 7.4952807869711675,
+      "learning_rate": 7.624789054109362e-06,
+      "loss": 0.71651783,
+      "memory(GiB)": 33.01,
+      "step": 450,
+      "train_speed(iter/s)": 0.191123
+    },
+    {
+      "acc": 0.86040745,
+      "epoch": 0.15076209410205435,
+      "grad_norm": 3.9434476633183064,
+      "learning_rate": 7.63858006009885e-06,
+      "loss": 0.67413173,
+      "memory(GiB)": 33.01,
+      "step": 455,
+      "train_speed(iter/s)": 0.191135
+    },
+    {
+      "acc": 0.86539726,
+      "epoch": 0.15241882041086813,
+      "grad_norm": 3.5118401097305174,
+      "learning_rate": 7.652220341694333e-06,
+      "loss": 0.72222414,
+      "memory(GiB)": 33.01,
+      "step": 460,
+      "train_speed(iter/s)": 0.191152
+    },
+    {
+      "acc": 0.85555248,
+      "epoch": 0.1540755467196819,
+      "grad_norm": 5.947399321976254,
+      "learning_rate": 7.6657131578979e-06,
+      "loss": 0.74063187,
+      "memory(GiB)": 33.01,
+      "step": 465,
+      "train_speed(iter/s)": 0.19117
+    },
+    {
+      "acc": 0.84811583,
+      "epoch": 0.15573227302849568,
+      "grad_norm": 5.697052687978797,
+      "learning_rate": 7.67906166314067e-06,
+      "loss": 0.77786217,
+      "memory(GiB)": 33.01,
+      "step": 470,
+      "train_speed(iter/s)": 0.191183
+    },
+    {
+      "acc": 0.86034212,
+      "epoch": 0.1573889993373095,
+      "grad_norm": 4.0454814939907875,
+      "learning_rate": 7.6922689117093e-06,
+      "loss": 0.69349651,
+      "memory(GiB)": 33.01,
+      "step": 475,
+      "train_speed(iter/s)": 0.191194
+    },
+    {
+      "acc": 0.85979271,
+      "epoch": 0.15904572564612326,
+      "grad_norm": 3.7930235007194,
+      "learning_rate": 7.7053378619407e-06,
+      "loss": 0.70015211,
+      "memory(GiB)": 33.01,
+      "step": 480,
+      "train_speed(iter/s)": 0.191206
+    },
+    {
+      "acc": 0.85328636,
+      "epoch": 0.16070245195493704,
+      "grad_norm": 7.599227481118352,
+      "learning_rate": 7.718271380199416e-06,
+      "loss": 0.69486294,
+      "memory(GiB)": 33.01,
+      "step": 485,
+      "train_speed(iter/s)": 0.191223
+    },
+    {
+      "acc": 0.85639191,
+      "epoch": 0.16235917826375082,
+      "grad_norm": 11.596092389773954,
+      "learning_rate": 7.73107224465102e-06,
+      "loss": 0.70257854,
+      "memory(GiB)": 33.01,
+      "step": 490,
+      "train_speed(iter/s)": 0.191238
+    },
+    {
+      "acc": 0.84910851,
+      "epoch": 0.16401590457256462,
+      "grad_norm": 9.20624588132978,
+      "learning_rate": 7.743743148843951e-06,
+      "loss": 0.81106377,
+      "memory(GiB)": 33.01,
+      "step": 495,
+      "train_speed(iter/s)": 0.191251
+    },
+    {
+      "acc": 0.84651041,
+      "epoch": 0.1656726308813784,
+      "grad_norm": 6.889524320452217,
+      "learning_rate": 7.756286705111353e-06,
+      "loss": 0.74286199,
+      "memory(GiB)": 33.01,
+      "step": 500,
+      "train_speed(iter/s)": 0.191263
+    },
+    {
+      "acc": 0.84956284,
+      "epoch": 0.16732935719019218,
+      "grad_norm": 5.966460531077995,
+      "learning_rate": 7.768705447803669e-06,
+      "loss": 0.74496927,
+      "memory(GiB)": 33.01,
+      "step": 505,
+      "train_speed(iter/s)": 0.191279
+    },
+    {
+      "acc": 0.84233456,
+      "epoch": 0.16898608349900596,
+      "grad_norm": 4.027572019536729,
+      "learning_rate": 7.781001836361969e-06,
+      "loss": 0.75807991,
+      "memory(GiB)": 33.01,
+      "step": 510,
+      "train_speed(iter/s)": 0.191293
+    },
+    {
+      "acc": 0.83378162,
+      "epoch": 0.17064280980781976,
+      "grad_norm": 3.5102186121946226,
+      "learning_rate": 7.79317825824136e-06,
+      "loss": 0.77981734,
+      "memory(GiB)": 33.01,
+      "step": 515,
+      "train_speed(iter/s)": 0.191308
+    },
+    {
+      "acc": 0.84073734,
+      "epoch": 0.17229953611663354,
+      "grad_norm": 5.853079987918349,
+      "learning_rate": 7.805237031693173e-06,
+      "loss": 0.78628778,
+      "memory(GiB)": 33.01,
+      "step": 520,
+      "train_speed(iter/s)": 0.19132
+    },
+    {
+      "acc": 0.84980087,
+      "epoch": 0.17395626242544732,
+      "grad_norm": 3.5454605857696357,
+      "learning_rate": 7.817180408414004e-06,
+      "loss": 0.76987505,
+      "memory(GiB)": 33.01,
+      "step": 525,
+      "train_speed(iter/s)": 0.191332
+    },
+    {
+      "acc": 0.84348345,
+      "epoch": 0.1756129887342611,
+      "grad_norm": 4.598681189813219,
+      "learning_rate": 7.829010576069216e-06,
+      "loss": 0.73552933,
+      "memory(GiB)": 33.01,
+      "step": 530,
+      "train_speed(iter/s)": 0.191345
+    },
+    {
+      "acc": 0.85787067,
+      "epoch": 0.17726971504307487,
+      "grad_norm": 4.604010264001306,
+      "learning_rate": 7.840729660697963e-06,
+      "loss": 0.71789432,
+      "memory(GiB)": 33.01,
+      "step": 535,
+      "train_speed(iter/s)": 0.191361
+    },
+    {
+      "acc": 0.84588737,
+      "epoch": 0.17892644135188868,
+      "grad_norm": 4.830877787847248,
+      "learning_rate": 7.852339729006332e-06,
+      "loss": 0.74034619,
+      "memory(GiB)": 33.01,
+      "step": 540,
+      "train_speed(iter/s)": 0.191371
+    },
+    {
+      "acc": 0.85173817,
+      "epoch": 0.18058316766070245,
+      "grad_norm": 2.8699572644791473,
+      "learning_rate": 7.863842790554826e-06,
+      "loss": 0.72552185,
+      "memory(GiB)": 33.01,
+      "step": 545,
+      "train_speed(iter/s)": 0.191382
+    },
+    {
+      "acc": 0.86452513,
+      "epoch": 0.18223989396951623,
+      "grad_norm": 7.246596519208709,
+      "learning_rate": 7.875240799845943e-06,
+      "loss": 0.74848604,
+      "memory(GiB)": 33.01,
+      "step": 550,
+      "train_speed(iter/s)": 0.191393
+    },
+    {
+      "acc": 0.8572896,
+      "epoch": 0.18389662027833,
+      "grad_norm": 3.7520051139701445,
+      "learning_rate": 7.886535658317292e-06,
+      "loss": 0.71552725,
+      "memory(GiB)": 33.01,
+      "step": 555,
+      "train_speed(iter/s)": 0.191405
+    },
+    {
+      "acc": 0.86681576,
+      "epoch": 0.1855533465871438,
+      "grad_norm": 3.5009893983455798,
+      "learning_rate": 7.897729216245343e-06,
+      "loss": 0.69532685,
+      "memory(GiB)": 33.01,
+      "step": 560,
+      "train_speed(iter/s)": 0.191418
+    },
+    {
+      "acc": 0.85972776,
+      "epoch": 0.1872100728959576,
+      "grad_norm": 5.035058722986529,
+      "learning_rate": 7.908823274564535e-06,
+      "loss": 0.69615574,
+      "memory(GiB)": 33.01,
+      "step": 565,
+      "train_speed(iter/s)": 0.191428
+    },
+    {
+      "acc": 0.84751177,
+      "epoch": 0.18886679920477137,
+      "grad_norm": 9.317941270292748,
+      "learning_rate": 7.919819586606273e-06,
+      "loss": 0.76375556,
+      "memory(GiB)": 33.01,
+      "step": 570,
+      "train_speed(iter/s)": 0.191435
+    },
+    {
+      "acc": 0.85486727,
+      "epoch": 0.19052352551358515,
+      "grad_norm": 4.402628274507517,
+      "learning_rate": 7.930719859761959e-06,
+      "loss": 0.73269348,
+      "memory(GiB)": 33.01,
+      "step": 575,
+      "train_speed(iter/s)": 0.191444
+    },
+    {
+      "acc": 0.847686,
+      "epoch": 0.19218025182239895,
+      "grad_norm": 6.094219099288395,
+      "learning_rate": 7.94152575707405e-06,
+      "loss": 0.75301404,
+      "memory(GiB)": 33.01,
+      "step": 580,
+      "train_speed(iter/s)": 0.191453
+    },
+    {
+      "acc": 0.85120449,
+      "epoch": 0.19383697813121273,
+      "grad_norm": 7.83197612062828,
+      "learning_rate": 7.952238898758805e-06,
+      "loss": 0.73857994,
+      "memory(GiB)": 33.01,
+      "step": 585,
+      "train_speed(iter/s)": 0.191465
+    },
+    {
+      "acc": 0.86423912,
+      "epoch": 0.1954937044400265,
+      "grad_norm": 5.655476710354406,
+      "learning_rate": 7.962860863664248e-06,
+      "loss": 0.66899776,
+      "memory(GiB)": 33.01,
+      "step": 590,
+      "train_speed(iter/s)": 0.191477
+    },
+    {
+      "acc": 0.85700264,
+      "epoch": 0.19715043074884028,
+      "grad_norm": 7.116192961530635,
+      "learning_rate": 7.97339319066661e-06,
+      "loss": 0.72598734,
+      "memory(GiB)": 33.01,
+      "step": 595,
+      "train_speed(iter/s)": 0.191488
+    },
+    {
+      "acc": 0.86150017,
+      "epoch": 0.1988071570576541,
+      "grad_norm": 6.748827026936556,
+      "learning_rate": 7.983837380008324e-06,
+      "loss": 0.67497487,
+      "memory(GiB)": 33.01,
+      "step": 600,
+      "train_speed(iter/s)": 0.191498
+    },
+    {
+      "acc": 0.85967827,
+      "epoch": 0.20046388336646787,
+      "grad_norm": 10.998154682065033,
+      "learning_rate": 7.994194894580531e-06,
+      "loss": 0.68231859,
+      "memory(GiB)": 33.01,
+      "step": 605,
+      "train_speed(iter/s)": 0.191485
+    },
+    {
+      "acc": 0.85635414,
+      "epoch": 0.20212060967528164,
+      "grad_norm": 4.5305781096214615,
+      "learning_rate": 8.00446716115277e-06,
+      "loss": 0.70385208,
+      "memory(GiB)": 33.01,
+      "step": 610,
+      "train_speed(iter/s)": 0.191497
+    },
+    {
+      "acc": 0.86400633,
+      "epoch": 0.20377733598409542,
+      "grad_norm": 8.628073552208958,
+      "learning_rate": 8.014655571552501e-06,
+      "loss": 0.69599981,
+      "memory(GiB)": 33.01,
+      "step": 615,
+      "train_speed(iter/s)": 0.191488
+    },
+    {
+      "acc": 0.85632267,
+      "epoch": 0.20543406229290923,
+      "grad_norm": 4.386175166940569,
+      "learning_rate": 8.024761483796864e-06,
+      "loss": 0.73796282,
+      "memory(GiB)": 33.01,
+      "step": 620,
+      "train_speed(iter/s)": 0.191498
+    },
+    {
+      "acc": 0.85163746,
+      "epoch": 0.207090788601723,
+      "grad_norm": 4.350461930883511,
+      "learning_rate": 8.034786223178978e-06,
+      "loss": 0.71875296,
+      "memory(GiB)": 33.01,
+      "step": 625,
+      "train_speed(iter/s)": 0.191504
+    },
+    {
+      "acc": 0.84534607,
+      "epoch": 0.20874751491053678,
+      "grad_norm": 6.498345108224841,
+      "learning_rate": 8.044731083310975e-06,
+      "loss": 0.77620149,
+      "memory(GiB)": 33.01,
+      "step": 630,
+      "train_speed(iter/s)": 0.191515
+    },
+    {
+      "acc": 0.85141335,
+      "epoch": 0.21040424121935056,
+      "grad_norm": 5.992706293797647,
+      "learning_rate": 8.054597327125787e-06,
+      "loss": 0.74544926,
+      "memory(GiB)": 33.01,
+      "step": 635,
+      "train_speed(iter/s)": 0.191521
+    },
+    {
+      "acc": 0.85580578,
+      "epoch": 0.21206096752816433,
+      "grad_norm": 2.7314353914768557,
+      "learning_rate": 8.064386187839663e-06,
+      "loss": 0.69488735,
+      "memory(GiB)": 33.01,
+      "step": 640,
+      "train_speed(iter/s)": 0.191531
+    },
+    {
+      "acc": 0.86037788,
+      "epoch": 0.21371769383697814,
+      "grad_norm": 4.153344973196596,
+      "learning_rate": 8.074098869877223e-06,
+      "loss": 0.69857273,
+      "memory(GiB)": 33.01,
+      "step": 645,
+      "train_speed(iter/s)": 0.19154
+    },
+    {
+      "acc": 0.85764399,
+      "epoch": 0.21537442014579192,
+      "grad_norm": 4.729778970937635,
+      "learning_rate": 8.083736549760797e-06,
+      "loss": 0.69933314,
+      "memory(GiB)": 33.01,
+      "step": 650,
+      "train_speed(iter/s)": 0.191548
+    },
+    {
+      "acc": 0.86113453,
+      "epoch": 0.2170311464546057,
+      "grad_norm": 5.008229766391558,
+      "learning_rate": 8.093300376965687e-06,
+      "loss": 0.72642188,
+      "memory(GiB)": 33.01,
+      "step": 655,
+      "train_speed(iter/s)": 0.191556
+    },
+    {
+      "acc": 0.85428047,
+      "epoch": 0.21868787276341947,
+      "grad_norm": 4.990582646367713,
+      "learning_rate": 8.102791474742914e-06,
+      "loss": 0.70296922,
+      "memory(GiB)": 33.01,
+      "step": 660,
+      "train_speed(iter/s)": 0.191562
+    },
+    {
+      "acc": 0.8543005,
+      "epoch": 0.22034459907223328,
+      "grad_norm": 4.814734543712982,
+      "learning_rate": 8.112210940910912e-06,
+      "loss": 0.7326674,
+      "memory(GiB)": 33.01,
+      "step": 665,
+      "train_speed(iter/s)": 0.191571
+    },
+    {
+      "acc": 0.87087936,
+      "epoch": 0.22200132538104705,
+      "grad_norm": 4.274810521989821,
+      "learning_rate": 8.121559848617575e-06,
+      "loss": 0.68122473,
+      "memory(GiB)": 33.01,
+      "step": 670,
+      "train_speed(iter/s)": 0.191583
+    },
+    {
+      "acc": 0.86698112,
+      "epoch": 0.22365805168986083,
+      "grad_norm": 4.626368782095732,
+      "learning_rate": 8.130839247073958e-06,
+      "loss": 0.67305284,
+      "memory(GiB)": 33.01,
+      "step": 675,
+      "train_speed(iter/s)": 0.191592
+    },
+    {
+      "acc": 0.8643259,
+      "epoch": 0.2253147779986746,
+      "grad_norm": 6.924801505180445,
+      "learning_rate": 8.140050162260932e-06,
+      "loss": 0.6990859,
+      "memory(GiB)": 33.01,
+      "step": 680,
+      "train_speed(iter/s)": 0.191601
+    },
+    {
+      "acc": 0.86108046,
+      "epoch": 0.22697150430748841,
+      "grad_norm": 5.728666374981688,
+      "learning_rate": 8.149193597609926e-06,
+      "loss": 0.73946257,
+      "memory(GiB)": 33.01,
+      "step": 685,
+      "train_speed(iter/s)": 0.191606
+    },
+    {
+      "acc": 0.86837921,
+      "epoch": 0.2286282306163022,
+      "grad_norm": 4.343056660340748,
+      "learning_rate": 8.15827053465893e-06,
+      "loss": 0.7034698,
+      "memory(GiB)": 33.01,
+      "step": 690,
+      "train_speed(iter/s)": 0.191613
+    },
+    {
+      "acc": 0.86654921,
+      "epoch": 0.23028495692511597,
+      "grad_norm": 5.829114137408897,
+      "learning_rate": 8.167281933684813e-06,
+      "loss": 0.67937307,
+      "memory(GiB)": 33.01,
+      "step": 695,
+      "train_speed(iter/s)": 0.191621
+    },
+    {
+      "acc": 0.86075773,
+      "epoch": 0.23194168323392975,
+      "grad_norm": 4.078790318510965,
+      "learning_rate": 8.176228734312967e-06,
+      "loss": 0.68860335,
+      "memory(GiB)": 33.01,
+      "step": 700,
+      "train_speed(iter/s)": 0.191629
+    },
+    {
+      "acc": 0.84542274,
+      "epoch": 0.23359840954274355,
+      "grad_norm": 10.29612897068814,
+      "learning_rate": 8.185111856105267e-06,
+      "loss": 0.76998425,
+      "memory(GiB)": 33.01,
+      "step": 705,
+      "train_speed(iter/s)": 0.191637
+    },
+    {
+      "acc": 0.85347729,
+      "epoch": 0.23525513585155733,
+      "grad_norm": 5.772386392955135,
+      "learning_rate": 8.19393219912725e-06,
+      "loss": 0.73821373,
+      "memory(GiB)": 33.01,
+      "step": 710,
+      "train_speed(iter/s)": 0.191648
+    },
+    {
+      "acc": 0.86139927,
+      "epoch": 0.2369118621603711,
+      "grad_norm": 9.905356064900213,
+      "learning_rate": 8.202690644495385e-06,
+      "loss": 0.74055252,
+      "memory(GiB)": 33.01,
+      "step": 715,
+      "train_speed(iter/s)": 0.191653
+    },
+    {
+      "acc": 0.85751944,
+      "epoch": 0.23856858846918488,
+      "grad_norm": 3.526084318998935,
+      "learning_rate": 8.211388054905297e-06,
+      "loss": 0.68814349,
+      "memory(GiB)": 33.01,
+      "step": 720,
+      "train_speed(iter/s)": 0.191662
+    },
+    {
+      "acc": 0.85765324,
+      "epoch": 0.24022531477799866,
+      "grad_norm": 4.954944455850853,
+      "learning_rate": 8.220025275141674e-06,
+      "loss": 0.65789051,
+      "memory(GiB)": 33.01,
+      "step": 725,
+      "train_speed(iter/s)": 0.19167
+    },
+    {
+      "acc": 0.8511692,
+      "epoch": 0.24188204108681247,
+      "grad_norm": 4.155498568727979,
+      "learning_rate": 8.228603132570695e-06,
+      "loss": 0.73796372,
+      "memory(GiB)": 33.01,
+      "step": 730,
+      "train_speed(iter/s)": 0.191676
+    },
+    {
+      "acc": 0.86511383,
+      "epoch": 0.24353876739562624,
+      "grad_norm": 4.69163212202949,
+      "learning_rate": 8.237122437615616e-06,
+      "loss": 0.69170351,
+      "memory(GiB)": 33.01,
+      "step": 735,
+      "train_speed(iter/s)": 0.191682
+    },
+    {
+      "acc": 0.86455679,
+      "epoch": 0.24519549370444002,
+      "grad_norm": 6.091645099912835,
+      "learning_rate": 8.245583984216255e-06,
+      "loss": 0.66270704,
+      "memory(GiB)": 33.01,
+      "step": 740,
+      "train_speed(iter/s)": 0.191691
+    },
+    {
+      "acc": 0.86024208,
+      "epoch": 0.2468522200132538,
+      "grad_norm": 4.783219472099627,
+      "learning_rate": 8.253988550272997e-06,
+      "loss": 0.71306019,
+      "memory(GiB)": 33.01,
+      "step": 745,
+      "train_speed(iter/s)": 0.1917
+    },
+    {
+      "acc": 0.86008968,
+      "epoch": 0.2485089463220676,
+      "grad_norm": 8.073864065552426,
+      "learning_rate": 8.26233689807595e-06,
+      "loss": 0.7465075,
+      "memory(GiB)": 33.01,
+      "step": 750,
+      "train_speed(iter/s)": 0.191707
+    },
+    {
+      "acc": 0.86828327,
+      "epoch": 0.25016567263088135,
+      "grad_norm": 3.2745862282999,
+      "learning_rate": 8.270629774719843e-06,
+      "loss": 0.64948173,
+      "memory(GiB)": 33.01,
+      "step": 755,
+      "train_speed(iter/s)": 0.191715
+    },
+    {
+      "acc": 0.86992846,
+      "epoch": 0.25182239893969516,
+      "grad_norm": 4.671054926219876,
+      "learning_rate": 8.278867912505236e-06,
+      "loss": 0.7060751,
+      "memory(GiB)": 33.01,
+      "step": 760,
+      "train_speed(iter/s)": 0.191724
+    },
+    {
+      "acc": 0.86491318,
+      "epoch": 0.25347912524850896,
+      "grad_norm": 4.807854418287031,
+      "learning_rate": 8.287052029326564e-06,
+      "loss": 0.69029131,
+      "memory(GiB)": 33.01,
+      "step": 765,
+      "train_speed(iter/s)": 0.191733
+    },
+    {
+      "acc": 0.8653554,
+      "epoch": 0.2551358515573227,
+      "grad_norm": 7.960100790874098,
+      "learning_rate": 8.295182829047555e-06,
+      "loss": 0.75642958,
+      "memory(GiB)": 33.01,
+      "step": 770,
+      "train_speed(iter/s)": 0.19174
+    },
+    {
+      "acc": 0.87609472,
+      "epoch": 0.2567925778661365,
+      "grad_norm": 3.7332405404403177,
+      "learning_rate": 8.30326100186449e-06,
+      "loss": 0.63245564,
+      "memory(GiB)": 33.01,
+      "step": 775,
+      "train_speed(iter/s)": 0.191746
+    },
+    {
+      "acc": 0.86726351,
+      "epoch": 0.2584493041749503,
+      "grad_norm": 6.592949946816494,
+      "learning_rate": 8.31128722465777e-06,
+      "loss": 0.70761013,
+      "memory(GiB)": 33.01,
+      "step": 780,
+      "train_speed(iter/s)": 0.191751
+    },
+    {
+      "acc": 0.85083599,
+      "epoch": 0.2601060304837641,
+      "grad_norm": 7.117368440138149,
+      "learning_rate": 8.319262161332283e-06,
+      "loss": 0.79634342,
+      "memory(GiB)": 33.01,
+      "step": 785,
+      "train_speed(iter/s)": 0.191757
+    },
+    {
+      "acc": 0.85279007,
+      "epoch": 0.2617627567925779,
+      "grad_norm": 5.0439827198012415,
+      "learning_rate": 8.32718646314694e-06,
+      "loss": 0.75035219,
+      "memory(GiB)": 33.01,
+      "step": 790,
+      "train_speed(iter/s)": 0.191765
+    },
+    {
+      "acc": 0.85604877,
+      "epoch": 0.2634194831013916,
+      "grad_norm": 3.5927726013912524,
+      "learning_rate": 8.335060769033813e-06,
+      "loss": 0.68453722,
+      "memory(GiB)": 33.01,
+      "step": 795,
+      "train_speed(iter/s)": 0.191771
+    },
+    {
+      "acc": 0.8636591,
+      "epoch": 0.26507620941020543,
+      "grad_norm": 3.8977563793690533,
+      "learning_rate": 8.342885705907289e-06,
+      "loss": 0.71641669,
+      "memory(GiB)": 33.01,
+      "step": 800,
+      "train_speed(iter/s)": 0.191778
+    },
+    {
+      "acc": 0.8587285,
+      "epoch": 0.26673293571901924,
+      "grad_norm": 3.5022622627110644,
+      "learning_rate": 8.350661888963573e-06,
+      "loss": 0.70699081,
+      "memory(GiB)": 33.01,
+      "step": 805,
+      "train_speed(iter/s)": 0.191784
+    },
+    {
+      "acc": 0.86881943,
+      "epoch": 0.268389662027833,
+      "grad_norm": 5.7848711558388155,
+      "learning_rate": 8.358389921970929e-06,
+      "loss": 0.67217498,
+      "memory(GiB)": 33.01,
+      "step": 810,
+      "train_speed(iter/s)": 0.191789
+    },
+    {
+      "acc": 0.86003571,
+      "epoch": 0.2700463883366468,
+      "grad_norm": 5.597400102240786,
+      "learning_rate": 8.366070397550993e-06,
+      "loss": 0.68687806,
+      "memory(GiB)": 33.01,
+      "step": 815,
+      "train_speed(iter/s)": 0.191794
+    },
+    {
+      "acc": 0.85528393,
+      "epoch": 0.2717031146454606,
+      "grad_norm": 5.648428195440093,
+      "learning_rate": 8.373703897451464e-06,
+      "loss": 0.70997171,
+      "memory(GiB)": 33.01,
+      "step": 820,
+      "train_speed(iter/s)": 0.191799
+    },
+    {
+      "acc": 0.86729469,
+      "epoch": 0.27335984095427435,
+      "grad_norm": 12.497625345993981,
+      "learning_rate": 8.38129099281054e-06,
+      "loss": 0.6821249,
+      "memory(GiB)": 33.01,
+      "step": 825,
+      "train_speed(iter/s)": 0.191803
+    },
+    {
+      "acc": 0.8640686,
+      "epoch": 0.27501656726308815,
+      "grad_norm": 3.8320240234117833,
+      "learning_rate": 8.38883224441332e-06,
+      "loss": 0.66615601,
+      "memory(GiB)": 33.01,
+      "step": 830,
+      "train_speed(iter/s)": 0.191811
+    },
+    {
+      "acc": 0.86246319,
+      "epoch": 0.2766732935719019,
+      "grad_norm": 5.522117887322051,
+      "learning_rate": 8.396328202940556e-06,
+      "loss": 0.70188322,
+      "memory(GiB)": 33.01,
+      "step": 835,
+      "train_speed(iter/s)": 0.191789
+    },
+    {
+      "acc": 0.86381741,
+      "epoch": 0.2783300198807157,
+      "grad_norm": 4.87827516864088,
+      "learning_rate": 8.403779409209938e-06,
+      "loss": 0.70778923,
+      "memory(GiB)": 33.01,
+      "step": 840,
+      "train_speed(iter/s)": 0.191797
+    },
+    {
+      "acc": 0.86391449,
+      "epoch": 0.2799867461895295,
+      "grad_norm": 6.395734537229641,
+      "learning_rate": 8.41118639441024e-06,
+      "loss": 0.68261509,
+      "memory(GiB)": 33.01,
+      "step": 845,
+      "train_speed(iter/s)": 0.191789
+    },
+    {
+      "acc": 0.85789928,
+      "epoch": 0.28164347249834326,
+      "grad_norm": 4.5005658378847775,
+      "learning_rate": 8.418549680328555e-06,
+      "loss": 0.71203232,
+      "memory(GiB)": 33.01,
+      "step": 850,
+      "train_speed(iter/s)": 0.191793
+    },
+    {
+      "acc": 0.84966297,
+      "epoch": 0.28330019880715707,
+      "grad_norm": 6.821844315190172,
+      "learning_rate": 8.425869779570868e-06,
+      "loss": 0.73459692,
+      "memory(GiB)": 33.01,
+      "step": 855,
+      "train_speed(iter/s)": 0.191797
+    },
+    {
+      "acc": 0.8597559,
+      "epoch": 0.2849569251159708,
+      "grad_norm": 8.223348678276244,
+      "learning_rate": 8.433147195776186e-06,
+      "loss": 0.70681105,
+      "memory(GiB)": 33.01,
+      "step": 860,
+      "train_speed(iter/s)": 0.191804
+    },
+    {
+      "acc": 0.86295443,
+      "epoch": 0.2866136514247846,
+      "grad_norm": 4.871278284312656,
+      "learning_rate": 8.440382423824487e-06,
+      "loss": 0.69193487,
+      "memory(GiB)": 33.01,
+      "step": 865,
+      "train_speed(iter/s)": 0.19181
+    },
+    {
+      "acc": 0.86210175,
+      "epoch": 0.2882703777335984,
+      "grad_norm": 6.510180490620472,
+      "learning_rate": 8.447575950038647e-06,
+      "loss": 0.71838708,
+      "memory(GiB)": 33.01,
+      "step": 870,
+      "train_speed(iter/s)": 0.191816
+    },
+    {
+      "acc": 0.86748362,
+      "epoch": 0.2899271040424122,
+      "grad_norm": 3.9864645425046445,
+      "learning_rate": 8.45472825238059e-06,
+      "loss": 0.70167303,
+      "memory(GiB)": 33.01,
+      "step": 875,
+      "train_speed(iter/s)": 0.191821
+    },
+    {
+      "acc": 0.84462318,
+      "epoch": 0.291583830351226,
+      "grad_norm": 8.047064350711658,
+      "learning_rate": 8.461839800641877e-06,
+      "loss": 0.78728943,
+      "memory(GiB)": 33.01,
+      "step": 880,
+      "train_speed(iter/s)": 0.191829
+    },
+    {
+      "acc": 0.85642376,
+      "epoch": 0.2932405566600398,
+      "grad_norm": 7.117747216483004,
+      "learning_rate": 8.468911056628845e-06,
+      "loss": 0.72243757,
+      "memory(GiB)": 33.01,
+      "step": 885,
+      "train_speed(iter/s)": 0.191835
+    },
+    {
+      "acc": 0.85076056,
+      "epoch": 0.29489728296885354,
+      "grad_norm": 7.759588962424344,
+      "learning_rate": 8.475942474342575e-06,
+      "loss": 0.72990608,
+      "memory(GiB)": 33.01,
+      "step": 890,
+      "train_speed(iter/s)": 0.191839
+    },
+    {
+      "acc": 0.86660328,
+      "epoch": 0.29655400927766734,
+      "grad_norm": 8.99467532941417,
+      "learning_rate": 8.482934500153795e-06,
+      "loss": 0.71121721,
+      "memory(GiB)": 33.01,
+      "step": 895,
+      "train_speed(iter/s)": 0.191845
+    },
+    {
+      "acc": 0.86047487,
+      "epoch": 0.2982107355864811,
+      "grad_norm": 4.933016923281054,
+      "learning_rate": 8.48988757297292e-06,
+      "loss": 0.69261165,
+      "memory(GiB)": 33.01,
+      "step": 900,
+      "train_speed(iter/s)": 0.191852
+    },
+    {
+      "acc": 0.85310354,
+      "epoch": 0.2998674618952949,
+      "grad_norm": 4.166546187502097,
+      "learning_rate": 8.496802124415386e-06,
+      "loss": 0.76197157,
+      "memory(GiB)": 33.01,
+      "step": 905,
+      "train_speed(iter/s)": 0.191858
+    },
+    {
+      "acc": 0.8576828,
+      "epoch": 0.3015241882041087,
+      "grad_norm": 5.2864064643943545,
+      "learning_rate": 8.50367857896241e-06,
+      "loss": 0.71239004,
+      "memory(GiB)": 33.01,
+      "step": 910,
+      "train_speed(iter/s)": 0.191864
+    },
+    {
+      "acc": 0.84969053,
+      "epoch": 0.30318091451292245,
+      "grad_norm": 3.9445977679759054,
+      "learning_rate": 8.510517354117366e-06,
+      "loss": 0.72859564,
+      "memory(GiB)": 33.01,
+      "step": 915,
+      "train_speed(iter/s)": 0.191868
+    },
+    {
+      "acc": 0.86889191,
+      "epoch": 0.30483764082173626,
+      "grad_norm": 5.859753689117367,
+      "learning_rate": 8.517318860557895e-06,
+      "loss": 0.69389181,
+      "memory(GiB)": 33.01,
+      "step": 920,
+      "train_speed(iter/s)": 0.191873
+    },
+    {
+      "acc": 0.85660534,
+      "epoch": 0.30649436713055006,
+      "grad_norm": 7.020475009305861,
+      "learning_rate": 8.52408350228388e-06,
+      "loss": 0.71614857,
+      "memory(GiB)": 33.01,
+      "step": 925,
+      "train_speed(iter/s)": 0.191878
+    },
+    {
+      "acc": 0.84998512,
+      "epoch": 0.3081510934393638,
+      "grad_norm": 5.157060283328341,
+      "learning_rate": 8.530811676761459e-06,
+      "loss": 0.73749046,
+      "memory(GiB)": 33.01,
+      "step": 930,
+      "train_speed(iter/s)": 0.191882
+    },
+    {
+      "acc": 0.84837933,
+      "epoch": 0.3098078197481776,
+      "grad_norm": 4.68265820623965,
+      "learning_rate": 8.537503775063145e-06,
+      "loss": 0.74418221,
+      "memory(GiB)": 33.01,
+      "step": 935,
+      "train_speed(iter/s)": 0.191887
+    },
+    {
+      "acc": 0.85448933,
+      "epoch": 0.31146454605699136,
+      "grad_norm": 4.407864788888527,
+      "learning_rate": 8.54416018200423e-06,
+      "loss": 0.71617479,
+      "memory(GiB)": 33.01,
+      "step": 940,
+      "train_speed(iter/s)": 0.191892
+    },
+    {
+      "acc": 0.86693363,
+      "epoch": 0.31312127236580517,
+      "grad_norm": 7.0917965746171046,
+      "learning_rate": 8.550781276275572e-06,
+      "loss": 0.68734121,
+      "memory(GiB)": 33.01,
+      "step": 945,
+      "train_speed(iter/s)": 0.191898
+    },
+    {
+      "acc": 0.85768795,
+      "epoch": 0.314777998674619,
+      "grad_norm": 5.484400495408929,
+      "learning_rate": 8.55736743057286e-06,
+      "loss": 0.7659287,
+      "memory(GiB)": 33.01,
+      "step": 950,
+      "train_speed(iter/s)": 0.191904
+    },
+    {
+      "acc": 0.87186584,
+      "epoch": 0.3164347249834327,
+      "grad_norm": 4.2330702758674565,
+      "learning_rate": 8.56391901172251e-06,
+      "loss": 0.66473894,
+      "memory(GiB)": 33.01,
+      "step": 955,
+      "train_speed(iter/s)": 0.19191
+    },
+    {
+      "acc": 0.85993299,
+      "epoch": 0.31809145129224653,
+      "grad_norm": 6.565656450893978,
+      "learning_rate": 8.57043638080426e-06,
+      "loss": 0.71878262,
+      "memory(GiB)": 33.01,
+      "step": 960,
+      "train_speed(iter/s)": 0.191913
+    },
+    {
+      "acc": 0.86838036,
+      "epoch": 0.3197481776010603,
+      "grad_norm": 5.553837202597705,
+      "learning_rate": 8.576919893270573e-06,
+      "loss": 0.71313562,
+      "memory(GiB)": 33.01,
+      "step": 965,
+      "train_speed(iter/s)": 0.191918
+    },
+    {
+      "acc": 0.85277061,
+      "epoch": 0.3214049039098741,
+      "grad_norm": 8.886485634732379,
+      "learning_rate": 8.583369899062976e-06,
+      "loss": 0.68710203,
+      "memory(GiB)": 33.01,
+      "step": 970,
+      "train_speed(iter/s)": 0.191921
+    },
+    {
+      "acc": 0.85232639,
+      "epoch": 0.3230616302186879,
+      "grad_norm": 7.470695382757026,
+      "learning_rate": 8.589786742725394e-06,
+      "loss": 0.70246749,
+      "memory(GiB)": 33.01,
+      "step": 975,
+      "train_speed(iter/s)": 0.191925
+    },
+    {
+      "acc": 0.85501528,
+      "epoch": 0.32471835652750164,
+      "grad_norm": 6.6965708100001615,
+      "learning_rate": 8.59617076351458e-06,
+      "loss": 0.73256102,
+      "memory(GiB)": 33.01,
+      "step": 980,
+      "train_speed(iter/s)": 0.191928
+    },
+    {
+      "acc": 0.85887146,
+      "epoch": 0.32637508283631544,
+      "grad_norm": 4.528870100887346,
+      "learning_rate": 8.602522295507758e-06,
+      "loss": 0.69201508,
+      "memory(GiB)": 33.01,
+      "step": 985,
+      "train_speed(iter/s)": 0.191933
+    },
+    {
+      "acc": 0.87829342,
+      "epoch": 0.32803180914512925,
+      "grad_norm": 4.286242120110182,
+      "learning_rate": 8.60884166770751e-06,
+      "loss": 0.64423809,
+      "memory(GiB)": 33.01,
+      "step": 990,
+      "train_speed(iter/s)": 0.191934
+    },
+    {
+      "acc": 0.87213535,
+      "epoch": 0.329688535453943,
+      "grad_norm": 4.327022130858467,
+      "learning_rate": 8.61512920414405e-06,
+      "loss": 0.68861742,
+      "memory(GiB)": 33.01,
+      "step": 995,
+      "train_speed(iter/s)": 0.191939
+    },
+    {
+      "acc": 0.86555452,
+      "epoch": 0.3313452617627568,
+      "grad_norm": 7.30777548783786,
+      "learning_rate": 8.621385223974913e-06,
+      "loss": 0.68824911,
+      "memory(GiB)": 33.01,
+      "step": 1000,
+      "train_speed(iter/s)": 0.191945
+    },
+    {
+      "acc": 0.84967728,
+      "epoch": 0.33300198807157055,
+      "grad_norm": 4.126697856691726,
+      "learning_rate": 8.627610041582171e-06,
+      "loss": 0.77748609,
+      "memory(GiB)": 33.01,
+      "step": 1005,
+      "train_speed(iter/s)": 0.19195
+    },
+    {
+      "acc": 0.84659004,
+      "epoch": 0.33465871438038436,
+      "grad_norm": 6.212526850023021,
+      "learning_rate": 8.633803966667229e-06,
+      "loss": 0.76965971,
+      "memory(GiB)": 33.01,
+      "step": 1010,
+      "train_speed(iter/s)": 0.191953
+    },
+    {
+      "acc": 0.84827509,
+      "epoch": 0.33631544068919816,
+      "grad_norm": 7.927500563444142,
+      "learning_rate": 8.639967304343287e-06,
+      "loss": 0.74794369,
+      "memory(GiB)": 33.01,
+      "step": 1015,
+      "train_speed(iter/s)": 0.191958
+    },
+    {
+      "acc": 0.85209866,
+      "epoch": 0.3379721669980119,
+      "grad_norm": 4.986422162591401,
+      "learning_rate": 8.646100355225527e-06,
+      "loss": 0.7509851,
+      "memory(GiB)": 33.01,
+      "step": 1020,
+      "train_speed(iter/s)": 0.191963
+    },
+    {
+      "acc": 0.86078835,
+      "epoch": 0.3396288933068257,
+      "grad_norm": 3.8140029467261494,
+      "learning_rate": 8.65220341551909e-06,
+      "loss": 0.70069637,
+      "memory(GiB)": 33.01,
+      "step": 1025,
+      "train_speed(iter/s)": 0.191969
+    },
+    {
+      "acc": 0.8566391,
+      "epoch": 0.3412856196156395,
+      "grad_norm": 2.553467166890378,
+      "learning_rate": 8.65827677710492e-06,
+      "loss": 0.74217367,
+      "memory(GiB)": 33.01,
+      "step": 1030,
+      "train_speed(iter/s)": 0.191972
+    },
+    {
+      "acc": 0.86919012,
+      "epoch": 0.3429423459244533,
+      "grad_norm": 2.7168711420548965,
+      "learning_rate": 8.664320727623527e-06,
+      "loss": 0.69090486,
+      "memory(GiB)": 33.01,
+      "step": 1035,
+      "train_speed(iter/s)": 0.191976
+    },
+    {
+      "acc": 0.86263781,
+      "epoch": 0.3445990722332671,
+      "grad_norm": 4.255977142857873,
+      "learning_rate": 8.670335550556733e-06,
+      "loss": 0.67670717,
+      "memory(GiB)": 33.01,
+      "step": 1040,
+      "train_speed(iter/s)": 0.191978
+    },
+    {
+      "acc": 0.86686478,
+      "epoch": 0.34625579854208083,
+      "grad_norm": 5.9461619636194,
+      "learning_rate": 8.676321525307448e-06,
+      "loss": 0.71811056,
+      "memory(GiB)": 33.01,
+      "step": 1045,
+      "train_speed(iter/s)": 0.191982
+    },
+    {
+      "acc": 0.86780119,
+      "epoch": 0.34791252485089463,
+      "grad_norm": 4.411280283281894,
+      "learning_rate": 8.682278927277564e-06,
+      "loss": 0.68573217,
+      "memory(GiB)": 33.01,
+      "step": 1050,
+      "train_speed(iter/s)": 0.191986
+    },
+    {
+      "acc": 0.86082306,
+      "epoch": 0.34956925115970844,
+      "grad_norm": 4.277181873250783,
+      "learning_rate": 8.688208027943965e-06,
+      "loss": 0.73803396,
+      "memory(GiB)": 33.01,
+      "step": 1055,
+      "train_speed(iter/s)": 0.19199
+    },
+    {
+      "acc": 0.85872288,
+      "epoch": 0.3512259774685222,
+      "grad_norm": 9.513126878642865,
+      "learning_rate": 8.694109094932776e-06,
+      "loss": 0.6754879,
+      "memory(GiB)": 33.01,
+      "step": 1060,
+      "train_speed(iter/s)": 0.19198
+    },
+    {
+      "acc": 0.87298717,
+      "epoch": 0.352882703777336,
+      "grad_norm": 6.519096238018569,
+      "learning_rate": 8.699982392091845e-06,
+      "loss": 0.64740787,
+      "memory(GiB)": 33.01,
+      "step": 1065,
+      "train_speed(iter/s)": 0.191983
+    },
+    {
+      "acc": 0.86751633,
+      "epoch": 0.35453943008614974,
+      "grad_norm": 5.5293755748294275,
+      "learning_rate": 8.705828179561522e-06,
+      "loss": 0.69447689,
+      "memory(GiB)": 33.01,
+      "step": 1070,
+      "train_speed(iter/s)": 0.191975
+    },
+    {
+      "acc": 0.8515645,
+      "epoch": 0.35619615639496355,
+      "grad_norm": 7.815726388251606,
+      "learning_rate": 8.711646713843812e-06,
+      "loss": 0.73291807,
+      "memory(GiB)": 33.01,
+      "step": 1075,
+      "train_speed(iter/s)": 0.191979
+    },
+    {
+      "acc": 0.85659885,
+      "epoch": 0.35785288270377735,
+      "grad_norm": 4.736123045402512,
+      "learning_rate": 8.717438247869894e-06,
+      "loss": 0.73431463,
+      "memory(GiB)": 33.01,
+      "step": 1080,
+      "train_speed(iter/s)": 0.191984
+    },
+    {
+      "acc": 0.86045341,
+      "epoch": 0.3595096090125911,
+      "grad_norm": 4.962469685077272,
+      "learning_rate": 8.723203031066102e-06,
+      "loss": 0.72274265,
+      "memory(GiB)": 33.01,
+      "step": 1085,
+      "train_speed(iter/s)": 0.191988
+    },
+    {
+      "acc": 0.86184435,
+      "epoch": 0.3611663353214049,
+      "grad_norm": 30.571469764149068,
+      "learning_rate": 8.728941309418385e-06,
+      "loss": 0.70956249,
+      "memory(GiB)": 33.01,
+      "step": 1090,
+      "train_speed(iter/s)": 0.191991
+    },
+    {
+      "acc": 0.8445303,
+      "epoch": 0.3628230616302187,
+      "grad_norm": 3.1331808301036417,
+      "learning_rate": 8.734653325535291e-06,
+      "loss": 0.75802627,
+      "memory(GiB)": 33.01,
+      "step": 1095,
+      "train_speed(iter/s)": 0.191993
+    },
+    {
+      "acc": 0.86017399,
+      "epoch": 0.36447978793903246,
+      "grad_norm": 6.255926580668658,
+      "learning_rate": 8.740339318709502e-06,
+      "loss": 0.70160847,
+      "memory(GiB)": 33.01,
+      "step": 1100,
+      "train_speed(iter/s)": 0.191998
+    },
+    {
+      "acc": 0.84508801,
+      "epoch": 0.36613651424784627,
+      "grad_norm": 9.187962899454137,
+      "learning_rate": 8.745999524978e-06,
+      "loss": 0.7189342,
+      "memory(GiB)": 33.01,
+      "step": 1105,
+      "train_speed(iter/s)": 0.192002
+    },
+    {
+      "acc": 0.87337513,
+      "epoch": 0.36779324055666,
+      "grad_norm": 3.6099000748388286,
+      "learning_rate": 8.751634177180852e-06,
+      "loss": 0.66577797,
+      "memory(GiB)": 33.01,
+      "step": 1110,
+      "train_speed(iter/s)": 0.192007
+    },
+    {
+      "acc": 0.85580883,
+      "epoch": 0.3694499668654738,
+      "grad_norm": 8.80790263065911,
+      "learning_rate": 8.757243505018693e-06,
+      "loss": 0.70440307,
+      "memory(GiB)": 33.01,
+      "step": 1115,
+      "train_speed(iter/s)": 0.19201
+    },
+    {
+      "acc": 0.84594364,
+      "epoch": 0.3711066931742876,
+      "grad_norm": 7.509362597486998,
+      "learning_rate": 8.762827735108903e-06,
+      "loss": 0.81703548,
+      "memory(GiB)": 33.01,
+      "step": 1120,
+      "train_speed(iter/s)": 0.192013
+    },
+    {
+      "acc": 0.86294937,
+      "epoch": 0.3727634194831014,
+      "grad_norm": 3.0748196059001103,
+      "learning_rate": 8.768387091040546e-06,
+      "loss": 0.7328989,
+      "memory(GiB)": 33.01,
+      "step": 1125,
+      "train_speed(iter/s)": 0.192018
+    },
+    {
+      "acc": 0.86343956,
+      "epoch": 0.3744201457919152,
+      "grad_norm": 3.956163390393462,
+      "learning_rate": 8.773921793428094e-06,
+      "loss": 0.72786388,
+      "memory(GiB)": 33.01,
+      "step": 1130,
+      "train_speed(iter/s)": 0.192022
+    },
+    {
+      "acc": 0.85795031,
+      "epoch": 0.376076872100729,
+      "grad_norm": 4.431043841633779,
+      "learning_rate": 8.779432059963951e-06,
+      "loss": 0.6969532,
+      "memory(GiB)": 33.01,
+      "step": 1135,
+      "train_speed(iter/s)": 0.192023
+    },
+    {
+      "acc": 0.8584589,
+      "epoch": 0.37773359840954274,
+      "grad_norm": 5.089581286774482,
+      "learning_rate": 8.784918105469833e-06,
+      "loss": 0.71613669,
+      "memory(GiB)": 33.01,
+      "step": 1140,
+      "train_speed(iter/s)": 0.192026
+    },
+    {
+      "acc": 0.86626835,
+      "epoch": 0.37939032471835654,
+      "grad_norm": 6.792031152265121,
+      "learning_rate": 8.79038014194702e-06,
+      "loss": 0.6801815,
+      "memory(GiB)": 33.01,
+      "step": 1145,
+      "train_speed(iter/s)": 0.192028
+    },
+    {
+      "acc": 0.86356316,
+      "epoch": 0.3810470510271703,
+      "grad_norm": 3.8554059385913884,
+      "learning_rate": 8.795818378625519e-06,
+      "loss": 0.6736372,
+      "memory(GiB)": 33.01,
+      "step": 1150,
+      "train_speed(iter/s)": 0.19203
+    },
+    {
+      "acc": 0.86050854,
+      "epoch": 0.3827037773359841,
+      "grad_norm": 3.2608560334191634,
+      "learning_rate": 8.801233022012152e-06,
+      "loss": 0.69969139,
+      "memory(GiB)": 33.01,
+      "step": 1155,
+      "train_speed(iter/s)": 0.192034
+    },
+    {
+      "acc": 0.85761337,
+      "epoch": 0.3843605036447979,
+      "grad_norm": 4.699570624072087,
+      "learning_rate": 8.80662427593761e-06,
+      "loss": 0.67253723,
+      "memory(GiB)": 33.01,
+      "step": 1160,
+      "train_speed(iter/s)": 0.192037
+    },
+    {
+      "acc": 0.8686121,
+      "epoch": 0.38601722995361165,
+      "grad_norm": 7.116748808234933,
+      "learning_rate": 8.811992341602492e-06,
+      "loss": 0.65382185,
+      "memory(GiB)": 33.01,
+      "step": 1165,
+      "train_speed(iter/s)": 0.192039
+    },
+    {
+      "acc": 0.87032061,
+      "epoch": 0.38767395626242546,
+      "grad_norm": 5.413912217697706,
+      "learning_rate": 8.817337417622365e-06,
+      "loss": 0.63192997,
+      "memory(GiB)": 33.01,
+      "step": 1170,
+      "train_speed(iter/s)": 0.192043
+    },
+    {
+      "acc": 0.86420956,
+      "epoch": 0.3893306825712392,
+      "grad_norm": 4.546166827237802,
+      "learning_rate": 8.822659700071856e-06,
+      "loss": 0.66633849,
+      "memory(GiB)": 33.01,
+      "step": 1175,
+      "train_speed(iter/s)": 0.192047
+    },
+    {
+      "acc": 0.86204653,
+      "epoch": 0.390987408880053,
+      "grad_norm": 6.9094640693737,
+      "learning_rate": 8.82795938252781e-06,
+      "loss": 0.69639988,
+      "memory(GiB)": 33.01,
+      "step": 1180,
+      "train_speed(iter/s)": 0.192051
+    },
+    {
+      "acc": 0.85785131,
+      "epoch": 0.3926441351888668,
+      "grad_norm": 5.6236949781970855,
+      "learning_rate": 8.833236656111535e-06,
+      "loss": 0.67771921,
+      "memory(GiB)": 33.01,
+      "step": 1185,
+      "train_speed(iter/s)": 0.192054
+    },
+    {
+      "acc": 0.86071796,
+      "epoch": 0.39430086149768057,
+      "grad_norm": 6.366038395999916,
+      "learning_rate": 8.83849170953017e-06,
+      "loss": 0.66028328,
+      "memory(GiB)": 33.01,
+      "step": 1190,
+      "train_speed(iter/s)": 0.192059
+    },
+    {
+      "acc": 0.87674227,
+      "epoch": 0.39595758780649437,
+      "grad_norm": 3.308309235543562,
+      "learning_rate": 8.843724729117152e-06,
+      "loss": 0.64831953,
+      "memory(GiB)": 33.01,
+      "step": 1195,
+      "train_speed(iter/s)": 0.192063
+    },
+    {
+      "acc": 0.86043034,
+      "epoch": 0.3976143141153082,
+      "grad_norm": 8.783944646097877,
+      "learning_rate": 8.848935898871885e-06,
+      "loss": 0.66111774,
+      "memory(GiB)": 33.01,
+      "step": 1200,
+      "train_speed(iter/s)": 0.192065
+    },
+    {
+      "acc": 0.85789833,
+      "epoch": 0.3992710404241219,
+      "grad_norm": 8.268745115458769,
+      "learning_rate": 8.854125400498542e-06,
+      "loss": 0.7165966,
+      "memory(GiB)": 33.01,
+      "step": 1205,
+      "train_speed(iter/s)": 0.192065
+    },
+    {
+      "acc": 0.85992565,
+      "epoch": 0.40092776673293573,
+      "grad_norm": 4.890343178737042,
+      "learning_rate": 8.85929341344409e-06,
+      "loss": 0.69850774,
+      "memory(GiB)": 33.01,
+      "step": 1210,
+      "train_speed(iter/s)": 0.192068
+    },
+    {
+      "acc": 0.86116114,
+      "epoch": 0.4025844930417495,
+      "grad_norm": 3.4696663332007764,
+      "learning_rate": 8.864440114935526e-06,
+      "loss": 0.67585459,
+      "memory(GiB)": 33.01,
+      "step": 1215,
+      "train_speed(iter/s)": 0.192071
+    },
+    {
+      "acc": 0.86599884,
+      "epoch": 0.4042412193505633,
+      "grad_norm": 3.400335100949336,
+      "learning_rate": 8.86956568001633e-06,
+      "loss": 0.65663977,
+      "memory(GiB)": 33.01,
+      "step": 1220,
+      "train_speed(iter/s)": 0.192074
+    },
+    {
+      "acc": 0.85257549,
+      "epoch": 0.4058979456593771,
+      "grad_norm": 4.291480298842552,
+      "learning_rate": 8.874670281582205e-06,
+      "loss": 0.70670052,
+      "memory(GiB)": 33.01,
+      "step": 1225,
+      "train_speed(iter/s)": 0.192078
+    },
+    {
+      "acc": 0.87359371,
+      "epoch": 0.40755467196819084,
+      "grad_norm": 6.877937879926714,
+      "learning_rate": 8.879754090416061e-06,
+      "loss": 0.64880548,
+      "memory(GiB)": 33.01,
+      "step": 1230,
+      "train_speed(iter/s)": 0.192079
+    },
+    {
+      "acc": 0.86227837,
+      "epoch": 0.40921139827700465,
+      "grad_norm": 5.624219130124923,
+      "learning_rate": 8.884817275222304e-06,
+      "loss": 0.68625078,
+      "memory(GiB)": 33.01,
+      "step": 1235,
+      "train_speed(iter/s)": 0.192083
+    },
+    {
+      "acc": 0.86328125,
+      "epoch": 0.41086812458581845,
+      "grad_norm": 5.297235302637378,
+      "learning_rate": 8.889860002660424e-06,
+      "loss": 0.68548312,
+      "memory(GiB)": 33.01,
+      "step": 1240,
+      "train_speed(iter/s)": 0.192086
+    },
+    {
+      "acc": 0.8626215,
+      "epoch": 0.4125248508946322,
+      "grad_norm": 6.606333534535208,
+      "learning_rate": 8.894882437377916e-06,
+      "loss": 0.68742762,
+      "memory(GiB)": 33.01,
+      "step": 1245,
+      "train_speed(iter/s)": 0.192089
+    },
+    {
+      "acc": 0.85593128,
+      "epoch": 0.414181577203446,
+      "grad_norm": 5.780530438267892,
+      "learning_rate": 8.899884742042536e-06,
+      "loss": 0.72823448,
+      "memory(GiB)": 33.01,
+      "step": 1250,
+      "train_speed(iter/s)": 0.192092
+    },
+    {
+      "acc": 0.84919424,
+      "epoch": 0.41583830351225975,
+      "grad_norm": 5.168353177298467,
+      "learning_rate": 8.904867077373915e-06,
+      "loss": 0.73082142,
+      "memory(GiB)": 33.01,
+      "step": 1255,
+      "train_speed(iter/s)": 0.192095
+    },
+    {
+      "acc": 0.86038704,
+      "epoch": 0.41749502982107356,
+      "grad_norm": 4.612606387706837,
+      "learning_rate": 8.909829602174535e-06,
+      "loss": 0.69704094,
+      "memory(GiB)": 33.01,
+      "step": 1260,
+      "train_speed(iter/s)": 0.192098
+    },
+    {
+      "acc": 0.86871624,
+      "epoch": 0.41915175612988737,
+      "grad_norm": 3.3570483389467007,
+      "learning_rate": 8.914772473360109e-06,
+      "loss": 0.70701017,
+      "memory(GiB)": 33.01,
+      "step": 1265,
+      "train_speed(iter/s)": 0.192102
+    },
+    {
+      "acc": 0.86309032,
+      "epoch": 0.4208084824387011,
+      "grad_norm": 3.2628703081995862,
+      "learning_rate": 8.919695845989347e-06,
+      "loss": 0.68439708,
+      "memory(GiB)": 33.01,
+      "step": 1270,
+      "train_speed(iter/s)": 0.192104
+    },
+    {
+      "acc": 0.87284012,
+      "epoch": 0.4224652087475149,
+      "grad_norm": 3.4484603804666825,
+      "learning_rate": 8.924599873293152e-06,
+      "loss": 0.65062599,
+      "memory(GiB)": 33.01,
+      "step": 1275,
+      "train_speed(iter/s)": 0.192106
+    },
+    {
+      "acc": 0.86791553,
+      "epoch": 0.42412193505632867,
+      "grad_norm": 8.973513100825011,
+      "learning_rate": 8.929484706703225e-06,
+      "loss": 0.67699113,
+      "memory(GiB)": 33.01,
+      "step": 1280,
+      "train_speed(iter/s)": 0.192109
+    },
+    {
+      "acc": 0.87207718,
+      "epoch": 0.4257786613651425,
+      "grad_norm": 6.517508560105469,
+      "learning_rate": 8.934350495880123e-06,
+      "loss": 0.64380426,
+      "memory(GiB)": 33.01,
+      "step": 1285,
+      "train_speed(iter/s)": 0.192111
+    },
+    {
+      "acc": 0.866152,
+      "epoch": 0.4274353876739563,
+      "grad_norm": 8.404196643915192,
+      "learning_rate": 8.939197388740783e-06,
+      "loss": 0.67536287,
+      "memory(GiB)": 33.01,
+      "step": 1290,
+      "train_speed(iter/s)": 0.192094
+    },
+    {
+      "acc": 0.87534895,
+      "epoch": 0.42909211398277003,
+      "grad_norm": 3.2016932440015915,
+      "learning_rate": 8.944025531485495e-06,
+      "loss": 0.63085623,
+      "memory(GiB)": 33.01,
+      "step": 1295,
+      "train_speed(iter/s)": 0.192098
+    },
+    {
+      "acc": 0.87227936,
+      "epoch": 0.43074884029158383,
+      "grad_norm": 3.8717486104123005,
+      "learning_rate": 8.948835068624357e-06,
+      "loss": 0.61535788,
+      "memory(GiB)": 33.01,
+      "step": 1300,
+      "train_speed(iter/s)": 0.192092
+    },
+    {
+      "acc": 0.86590223,
+      "epoch": 0.43240556660039764,
+      "grad_norm": 4.708375337715233,
+      "learning_rate": 8.953626143003244e-06,
+      "loss": 0.64982243,
+      "memory(GiB)": 33.01,
+      "step": 1305,
+      "train_speed(iter/s)": 0.192094
+    },
+    {
+      "acc": 0.879953,
+      "epoch": 0.4340622929092114,
+      "grad_norm": 3.8588317689639102,
+      "learning_rate": 8.958398895829247e-06,
+      "loss": 0.65139589,
+      "memory(GiB)": 33.01,
+      "step": 1310,
+      "train_speed(iter/s)": 0.192096
+    },
+    {
+      "acc": 0.87267056,
+      "epoch": 0.4357190192180252,
+      "grad_norm": 11.343379083443592,
+      "learning_rate": 8.963153466695663e-06,
+      "loss": 0.67646942,
+      "memory(GiB)": 33.01,
+      "step": 1315,
+      "train_speed(iter/s)": 0.1921
+    },
+    {
+      "acc": 0.8704361,
+      "epoch": 0.43737574552683894,
+      "grad_norm": 5.908780003246269,
+      "learning_rate": 8.967889993606474e-06,
+      "loss": 0.66506381,
+      "memory(GiB)": 33.01,
+      "step": 1320,
+      "train_speed(iter/s)": 0.192103
+    },
+    {
+      "acc": 0.87443924,
+      "epoch": 0.43903247183565275,
+      "grad_norm": 3.891369282847532,
+      "learning_rate": 8.972608613000402e-06,
+      "loss": 0.6349452,
+      "memory(GiB)": 33.01,
+      "step": 1325,
+      "train_speed(iter/s)": 0.192105
+    },
+    {
+      "acc": 0.86603394,
+      "epoch": 0.44068919814446655,
+      "grad_norm": 4.56074550696968,
+      "learning_rate": 8.977309459774474e-06,
+      "loss": 0.71010776,
+      "memory(GiB)": 33.01,
+      "step": 1330,
+      "train_speed(iter/s)": 0.192108
+    },
+    {
+      "acc": 0.86591244,
+      "epoch": 0.4423459244532803,
+      "grad_norm": 3.7422516690378202,
+      "learning_rate": 8.981992667307172e-06,
+      "loss": 0.66119232,
+      "memory(GiB)": 33.01,
+      "step": 1335,
+      "train_speed(iter/s)": 0.19211
+    },
+    {
+      "acc": 0.86200562,
+      "epoch": 0.4440026507620941,
+      "grad_norm": 5.2778052922519745,
+      "learning_rate": 8.986658367481134e-06,
+      "loss": 0.70098228,
+      "memory(GiB)": 33.01,
+      "step": 1340,
+      "train_speed(iter/s)": 0.192113
+    },
+    {
+      "acc": 0.86360703,
+      "epoch": 0.44565937707090786,
+      "grad_norm": 3.5249559587809194,
+      "learning_rate": 8.99130669070545e-06,
+      "loss": 0.68683853,
+      "memory(GiB)": 33.01,
+      "step": 1345,
+      "train_speed(iter/s)": 0.192116
+    },
+    {
+      "acc": 0.86272469,
+      "epoch": 0.44731610337972166,
+      "grad_norm": 6.0971293097221375,
+      "learning_rate": 8.995937765937517e-06,
+      "loss": 0.67999382,
+      "memory(GiB)": 33.01,
+      "step": 1350,
+      "train_speed(iter/s)": 0.192121
+    },
+    {
+      "acc": 0.86906853,
+      "epoch": 0.44897282968853547,
+      "grad_norm": 4.865638580202748,
+      "learning_rate": 9.00055172070452e-06,
+      "loss": 0.67824078,
+      "memory(GiB)": 33.01,
+      "step": 1355,
+      "train_speed(iter/s)": 0.192126
+    },
+    {
+      "acc": 0.8755106,
+      "epoch": 0.4506295559973492,
+      "grad_norm": 2.9266592704460512,
+      "learning_rate": 9.00514868112449e-06,
+      "loss": 0.64030924,
+      "memory(GiB)": 33.01,
+      "step": 1360,
+      "train_speed(iter/s)": 0.192129
+    },
+    {
+      "acc": 0.86524611,
+      "epoch": 0.452286282306163,
+      "grad_norm": 5.091998354549703,
+      "learning_rate": 9.009728771927006e-06,
+      "loss": 0.71764174,
+      "memory(GiB)": 33.01,
+      "step": 1365,
+      "train_speed(iter/s)": 0.192134
+    },
+    {
+      "acc": 0.86362028,
+      "epoch": 0.45394300861497683,
+      "grad_norm": 5.36536430236198,
+      "learning_rate": 9.014292116473486e-06,
+      "loss": 0.72176256,
+      "memory(GiB)": 33.01,
+      "step": 1370,
+      "train_speed(iter/s)": 0.192138
+    },
+    {
+      "acc": 0.85922489,
+      "epoch": 0.4555997349237906,
+      "grad_norm": 5.6030024883421925,
+      "learning_rate": 9.018838836777126e-06,
+      "loss": 0.70936112,
+      "memory(GiB)": 33.01,
+      "step": 1375,
+      "train_speed(iter/s)": 0.192141
+    },
+    {
+      "acc": 0.86305695,
+      "epoch": 0.4572564612326044,
+      "grad_norm": 3.3750177882476686,
+      "learning_rate": 9.02336905352249e-06,
+      "loss": 0.69946198,
+      "memory(GiB)": 33.01,
+      "step": 1380,
+      "train_speed(iter/s)": 0.192145
+    },
+    {
+      "acc": 0.85531149,
+      "epoch": 0.45891318754141813,
+      "grad_norm": 6.3626938674314575,
+      "learning_rate": 9.02788288608471e-06,
+      "loss": 0.71012573,
+      "memory(GiB)": 33.01,
+      "step": 1385,
+      "train_speed(iter/s)": 0.192148
+    },
+    {
+      "acc": 0.84457617,
+      "epoch": 0.46056991385023194,
+      "grad_norm": 5.522657483538765,
+      "learning_rate": 9.032380452548372e-06,
+      "loss": 0.72450757,
+      "memory(GiB)": 33.01,
+      "step": 1390,
+      "train_speed(iter/s)": 0.192151
+    },
+    {
+      "acc": 0.87537985,
+      "epoch": 0.46222664015904574,
+      "grad_norm": 3.9361085118086874,
+      "learning_rate": 9.036861869726056e-06,
+      "loss": 0.6987277,
+      "memory(GiB)": 33.01,
+      "step": 1395,
+      "train_speed(iter/s)": 0.192155
+    },
+    {
+      "acc": 0.86897469,
+      "epoch": 0.4638833664678595,
+      "grad_norm": 5.337820215319609,
+      "learning_rate": 9.041327253176527e-06,
+      "loss": 0.6750411,
+      "memory(GiB)": 33.01,
+      "step": 1400,
+      "train_speed(iter/s)": 0.192159
+    },
+    {
+      "acc": 0.85538292,
+      "epoch": 0.4655400927766733,
+      "grad_norm": 3.4327882607856943,
+      "learning_rate": 9.045776717222626e-06,
+      "loss": 0.7254981,
+      "memory(GiB)": 33.01,
+      "step": 1405,
+      "train_speed(iter/s)": 0.192163
+    },
+    {
+      "acc": 0.87039423,
+      "epoch": 0.4671968190854871,
+      "grad_norm": 11.219916917664618,
+      "learning_rate": 9.050210374968827e-06,
+      "loss": 0.67206554,
+      "memory(GiB)": 33.01,
+      "step": 1410,
+      "train_speed(iter/s)": 0.192166
+    },
+    {
+      "acc": 0.85317097,
+      "epoch": 0.46885354539430085,
+      "grad_norm": 12.478943442240555,
+      "learning_rate": 9.054628338318491e-06,
+      "loss": 0.75601158,
+      "memory(GiB)": 33.01,
+      "step": 1415,
+      "train_speed(iter/s)": 0.19217
+    },
+    {
+      "acc": 0.87292891,
+      "epoch": 0.47051027170311466,
+      "grad_norm": 4.570693229667621,
+      "learning_rate": 9.05903071799081e-06,
+      "loss": 0.64433074,
+      "memory(GiB)": 33.01,
+      "step": 1420,
+      "train_speed(iter/s)": 0.192172
+    },
+    {
+      "acc": 0.86060562,
+      "epoch": 0.4721669980119284,
+      "grad_norm": 3.9476140322917415,
+      "learning_rate": 9.063417623537456e-06,
+      "loss": 0.69826007,
+      "memory(GiB)": 33.01,
+      "step": 1425,
+      "train_speed(iter/s)": 0.192175
+    },
+    {
+      "acc": 0.87749691,
+      "epoch": 0.4738237243207422,
+      "grad_norm": 3.679397051485356,
+      "learning_rate": 9.067789163358945e-06,
+      "loss": 0.66564126,
+      "memory(GiB)": 33.01,
+      "step": 1430,
+      "train_speed(iter/s)": 0.192178
+    },
+    {
+      "acc": 0.86229973,
+      "epoch": 0.475480450629556,
+      "grad_norm": 3.8901355992289397,
+      "learning_rate": 9.072145444720704e-06,
+      "loss": 0.69365253,
+      "memory(GiB)": 33.01,
+      "step": 1435,
+      "train_speed(iter/s)": 0.192181
+    },
+    {
+      "acc": 0.86734676,
+      "epoch": 0.47713717693836977,
+      "grad_norm": 5.701611205590434,
+      "learning_rate": 9.076486573768857e-06,
+      "loss": 0.70805588,
+      "memory(GiB)": 33.01,
+      "step": 1440,
+      "train_speed(iter/s)": 0.192183
+    },
+    {
+      "acc": 0.86186886,
+      "epoch": 0.47879390324718357,
+      "grad_norm": 4.134391868806285,
+      "learning_rate": 9.080812655545758e-06,
+      "loss": 0.71403632,
+      "memory(GiB)": 33.01,
+      "step": 1445,
+      "train_speed(iter/s)": 0.192186
+    },
+    {
+      "acc": 0.85843849,
+      "epoch": 0.4804506295559973,
+      "grad_norm": 8.870580174999937,
+      "learning_rate": 9.085123794005234e-06,
+      "loss": 0.70811243,
+      "memory(GiB)": 33.01,
+      "step": 1450,
+      "train_speed(iter/s)": 0.192189
+    },
+    {
+      "acc": 0.86999588,
+      "epoch": 0.4821073558648111,
+      "grad_norm": 2.8181385446303358,
+      "learning_rate": 9.089420092027573e-06,
+      "loss": 0.62644167,
+      "memory(GiB)": 33.01,
+      "step": 1455,
+      "train_speed(iter/s)": 0.19219
+    },
+    {
+      "acc": 0.87824745,
+      "epoch": 0.48376408217362493,
+      "grad_norm": 3.2066696614980414,
+      "learning_rate": 9.093701651434256e-06,
+      "loss": 0.62812157,
+      "memory(GiB)": 33.01,
+      "step": 1460,
+      "train_speed(iter/s)": 0.192193
+    },
+    {
+      "acc": 0.87342119,
+      "epoch": 0.4854208084824387,
+      "grad_norm": 2.698069952587491,
+      "learning_rate": 9.097968573002436e-06,
+      "loss": 0.61367073,
+      "memory(GiB)": 33.01,
+      "step": 1465,
+      "train_speed(iter/s)": 0.192195
+    },
+    {
+      "acc": 0.87189951,
+      "epoch": 0.4870775347912525,
+      "grad_norm": 3.6121057523052764,
+      "learning_rate": 9.102220956479176e-06,
+      "loss": 0.60860167,
+      "memory(GiB)": 33.01,
+      "step": 1470,
+      "train_speed(iter/s)": 0.192198
+    },
+    {
+      "acc": 0.86669426,
+      "epoch": 0.4887342611000663,
+      "grad_norm": 7.7696093549689005,
+      "learning_rate": 9.106458900595433e-06,
+      "loss": 0.68489089,
+      "memory(GiB)": 33.01,
+      "step": 1475,
+      "train_speed(iter/s)": 0.192201
+    },
+    {
+      "acc": 0.87739582,
+      "epoch": 0.49039098740888004,
+      "grad_norm": 6.282902416033516,
+      "learning_rate": 9.110682503079815e-06,
+      "loss": 0.64681358,
+      "memory(GiB)": 33.01,
+      "step": 1480,
+      "train_speed(iter/s)": 0.192204
+    },
+    {
+      "acc": 0.87366724,
+      "epoch": 0.49204771371769385,
+      "grad_norm": 6.148845107605522,
+      "learning_rate": 9.114891860672107e-06,
+      "loss": 0.68622417,
+      "memory(GiB)": 33.01,
+      "step": 1485,
+      "train_speed(iter/s)": 0.192207
+    },
+    {
+      "acc": 0.86645737,
+      "epoch": 0.4937044400265076,
+      "grad_norm": 7.102791414384401,
+      "learning_rate": 9.119087069136557e-06,
+      "loss": 0.66596956,
+      "memory(GiB)": 33.01,
+      "step": 1490,
+      "train_speed(iter/s)": 0.19221
+    },
+    {
+      "acc": 0.87296562,
+      "epoch": 0.4953611663353214,
+      "grad_norm": 9.74166225216245,
+      "learning_rate": 9.123268223274961e-06,
+      "loss": 0.66302705,
+      "memory(GiB)": 33.01,
+      "step": 1495,
+      "train_speed(iter/s)": 0.192212
+    },
+    {
+      "acc": 0.86229172,
+      "epoch": 0.4970178926441352,
+      "grad_norm": 5.848311959392031,
+      "learning_rate": 9.12743541693951e-06,
+      "loss": 0.70333743,
+      "memory(GiB)": 33.01,
+      "step": 1500,
+      "train_speed(iter/s)": 0.192216
+    },
+    {
+      "acc": 0.87268381,
+      "epoch": 0.49867461895294896,
+      "grad_norm": 4.150118092077552,
+      "learning_rate": 9.131588743045426e-06,
+      "loss": 0.64809484,
+      "memory(GiB)": 33.01,
+      "step": 1505,
+      "train_speed(iter/s)": 0.192218
+    },
+    {
+      "acc": 0.87299833,
+      "epoch": 0.5003313452617627,
+      "grad_norm": 4.538142415377428,
+      "learning_rate": 9.135728293583403e-06,
+      "loss": 0.63944387,
+      "memory(GiB)": 33.01,
+      "step": 1510,
+      "train_speed(iter/s)": 0.192221
+    },
+    {
+      "acc": 0.87983656,
+      "epoch": 0.5019880715705766,
+      "grad_norm": 3.08163951023743,
+      "learning_rate": 9.139854159631825e-06,
+      "loss": 0.63588171,
+      "memory(GiB)": 33.01,
+      "step": 1515,
+      "train_speed(iter/s)": 0.192215
+    },
+    {
+      "acc": 0.86658287,
+      "epoch": 0.5036447978793903,
+      "grad_norm": 4.263188194458355,
+      "learning_rate": 9.143966431368796e-06,
+      "loss": 0.67643175,
+      "memory(GiB)": 33.01,
+      "step": 1520,
+      "train_speed(iter/s)": 0.19221
+    },
+    {
+      "acc": 0.87198019,
+      "epoch": 0.5053015241882041,
+      "grad_norm": 4.218782555945772,
+      "learning_rate": 9.148065198083954e-06,
+      "loss": 0.6854351,
+      "memory(GiB)": 33.01,
+      "step": 1525,
+      "train_speed(iter/s)": 0.192204
+    },
+    {
+      "acc": 0.86166058,
+      "epoch": 0.5069582504970179,
+      "grad_norm": 5.475988037012612,
+      "learning_rate": 9.152150548190123e-06,
+      "loss": 0.68347588,
+      "memory(GiB)": 33.01,
+      "step": 1530,
+      "train_speed(iter/s)": 0.192208
+    },
+    {
+      "acc": 0.86631327,
+      "epoch": 0.5086149768058317,
+      "grad_norm": 6.588801301856675,
+      "learning_rate": 9.156222569234739e-06,
+      "loss": 0.66437025,
+      "memory(GiB)": 33.01,
+      "step": 1535,
+      "train_speed(iter/s)": 0.19221
+    },
+    {
+      "acc": 0.87238407,
+      "epoch": 0.5102717031146454,
+      "grad_norm": 3.9675870595193046,
+      "learning_rate": 9.160281347911115e-06,
+      "loss": 0.67067642,
+      "memory(GiB)": 33.01,
+      "step": 1540,
+      "train_speed(iter/s)": 0.192214
+    },
+    {
+      "acc": 0.86192398,
+      "epoch": 0.5119284294234593,
+      "grad_norm": 6.149770568367035,
+      "learning_rate": 9.164326970069515e-06,
+      "loss": 0.67884417,
+      "memory(GiB)": 33.01,
+      "step": 1545,
+      "train_speed(iter/s)": 0.192218
+    },
+    {
+      "acc": 0.85227327,
+      "epoch": 0.513585155732273,
+      "grad_norm": 5.331320167629407,
+      "learning_rate": 9.16835952072805e-06,
+      "loss": 0.75046368,
+      "memory(GiB)": 33.01,
+      "step": 1550,
+      "train_speed(iter/s)": 0.192221
+    },
+    {
+      "acc": 0.85830784,
+      "epoch": 0.5152418820410868,
+      "grad_norm": 3.7113603011577325,
+      "learning_rate": 9.17237908408339e-06,
+      "loss": 0.68575492,
+      "memory(GiB)": 33.01,
+      "step": 1555,
+      "train_speed(iter/s)": 0.192224
+    },
+    {
+      "acc": 0.85954647,
+      "epoch": 0.5168986083499006,
+      "grad_norm": 4.575805064916573,
+      "learning_rate": 9.176385743521328e-06,
+      "loss": 0.67908902,
+      "memory(GiB)": 33.01,
+      "step": 1560,
+      "train_speed(iter/s)": 0.192227
+    },
+    {
+      "acc": 0.87961903,
+      "epoch": 0.5185553346587144,
+      "grad_norm": 2.5861178107326013,
+      "learning_rate": 9.180379581627149e-06,
+      "loss": 0.62059913,
+      "memory(GiB)": 33.01,
+      "step": 1565,
+      "train_speed(iter/s)": 0.19223
+    },
+    {
+      "acc": 0.87046461,
+      "epoch": 0.5202120609675281,
+      "grad_norm": 5.7201873315095275,
+      "learning_rate": 9.184360680195843e-06,
+      "loss": 0.65617762,
+      "memory(GiB)": 33.01,
+      "step": 1570,
+      "train_speed(iter/s)": 0.192233
+    },
+    {
+      "acc": 0.87465582,
+      "epoch": 0.5218687872763419,
+      "grad_norm": 4.51594779110544,
+      "learning_rate": 9.18832912024216e-06,
+      "loss": 0.64734449,
+      "memory(GiB)": 33.01,
+      "step": 1575,
+      "train_speed(iter/s)": 0.192235
+    },
+    {
+      "acc": 0.87109985,
+      "epoch": 0.5235255135851558,
+      "grad_norm": 7.116594105071289,
+      "learning_rate": 9.1922849820105e-06,
+      "loss": 0.63359661,
+      "memory(GiB)": 33.01,
+      "step": 1580,
+      "train_speed(iter/s)": 0.192238
+    },
+    {
+      "acc": 0.87022362,
+      "epoch": 0.5251822398939695,
+      "grad_norm": 6.455544574812186,
+      "learning_rate": 9.196228344984651e-06,
+      "loss": 0.63237476,
+      "memory(GiB)": 33.01,
+      "step": 1585,
+      "train_speed(iter/s)": 0.192241
+    },
+    {
+      "acc": 0.86301365,
+      "epoch": 0.5268389662027833,
+      "grad_norm": 8.113697111420223,
+      "learning_rate": 9.200159287897373e-06,
+      "loss": 0.68346348,
+      "memory(GiB)": 33.01,
+      "step": 1590,
+      "train_speed(iter/s)": 0.192245
+    },
+    {
+      "acc": 0.87428303,
+      "epoch": 0.5284956925115971,
+      "grad_norm": 5.0813744801911644,
+      "learning_rate": 9.204077888739822e-06,
+      "loss": 0.62634401,
+      "memory(GiB)": 33.01,
+      "step": 1595,
+      "train_speed(iter/s)": 0.192247
+    },
+    {
+      "acc": 0.88175039,
+      "epoch": 0.5301524188204109,
+      "grad_norm": 6.862745130743622,
+      "learning_rate": 9.207984224770848e-06,
+      "loss": 0.5954772,
+      "memory(GiB)": 33.01,
+      "step": 1600,
+      "train_speed(iter/s)": 0.19225
+    },
+    {
+      "acc": 0.88652163,
+      "epoch": 0.5318091451292246,
+      "grad_norm": 5.095229068290511,
+      "learning_rate": 9.211878372526119e-06,
+      "loss": 0.61880541,
+      "memory(GiB)": 33.01,
+      "step": 1605,
+      "train_speed(iter/s)": 0.192254
+    },
+    {
+      "acc": 0.88491421,
+      "epoch": 0.5334658714380385,
+      "grad_norm": 7.683127686889883,
+      "learning_rate": 9.215760407827133e-06,
+      "loss": 0.57899289,
+      "memory(GiB)": 33.01,
+      "step": 1610,
+      "train_speed(iter/s)": 0.192257
+    },
+    {
+      "acc": 0.87476616,
+      "epoch": 0.5351225977468522,
+      "grad_norm": 7.282543101591841,
+      "learning_rate": 9.219630405790062e-06,
+      "loss": 0.67257605,
+      "memory(GiB)": 33.01,
+      "step": 1615,
+      "train_speed(iter/s)": 0.192261
+    },
+    {
+      "acc": 0.86599264,
+      "epoch": 0.536779324055666,
+      "grad_norm": 8.087444973583551,
+      "learning_rate": 9.223488440834489e-06,
+      "loss": 0.70106878,
+      "memory(GiB)": 33.01,
+      "step": 1620,
+      "train_speed(iter/s)": 0.192266
+    },
+    {
+      "acc": 0.86528091,
+      "epoch": 0.5384360503644798,
+      "grad_norm": 9.746503473311424,
+      "learning_rate": 9.227334586691982e-06,
+      "loss": 0.72642713,
+      "memory(GiB)": 33.01,
+      "step": 1625,
+      "train_speed(iter/s)": 0.192268
+    },
+    {
+      "acc": 0.86779509,
+      "epoch": 0.5400927766732936,
+      "grad_norm": 8.615700205965208,
+      "learning_rate": 9.231168916414552e-06,
+      "loss": 0.65600066,
+      "memory(GiB)": 33.01,
+      "step": 1630,
+      "train_speed(iter/s)": 0.19227
+    },
+    {
+      "acc": 0.86514606,
+      "epoch": 0.5417495029821073,
+      "grad_norm": 5.225316406515859,
+      "learning_rate": 9.234991502382982e-06,
+      "loss": 0.65406122,
+      "memory(GiB)": 33.01,
+      "step": 1635,
+      "train_speed(iter/s)": 0.192274
+    },
+    {
+      "acc": 0.86649914,
+      "epoch": 0.5434062292909212,
+      "grad_norm": 3.340031154531065,
+      "learning_rate": 9.238802416315024e-06,
+      "loss": 0.66948366,
+      "memory(GiB)": 33.01,
+      "step": 1640,
+      "train_speed(iter/s)": 0.192277
+    },
+    {
+      "acc": 0.87296982,
+      "epoch": 0.5450629555997349,
+      "grad_norm": 4.8852907299692445,
+      "learning_rate": 9.242601729273468e-06,
+      "loss": 0.6427763,
+      "memory(GiB)": 33.01,
+      "step": 1645,
+      "train_speed(iter/s)": 0.19228
+    },
+    {
+      "acc": 0.86559639,
+      "epoch": 0.5467196819085487,
+      "grad_norm": 6.885611781964723,
+      "learning_rate": 9.2463895116741e-06,
+      "loss": 0.64991856,
+      "memory(GiB)": 33.01,
+      "step": 1650,
+      "train_speed(iter/s)": 0.192283
+    },
+    {
+      "acc": 0.85878677,
+      "epoch": 0.5483764082173624,
+      "grad_norm": 7.638042979861447,
+      "learning_rate": 9.250165833293521e-06,
+      "loss": 0.70569301,
+      "memory(GiB)": 33.01,
+      "step": 1655,
+      "train_speed(iter/s)": 0.192285
+    },
+    {
+      "acc": 0.87370911,
+      "epoch": 0.5500331345261763,
+      "grad_norm": 5.079730129726356,
+      "learning_rate": 9.25393076327688e-06,
+      "loss": 0.61256323,
+      "memory(GiB)": 33.01,
+      "step": 1660,
+      "train_speed(iter/s)": 0.192288
+    },
+    {
+      "acc": 0.87358227,
+      "epoch": 0.55168986083499,
+      "grad_norm": 4.056124455478207,
+      "learning_rate": 9.257684370145448e-06,
+      "loss": 0.61129341,
+      "memory(GiB)": 33.01,
+      "step": 1665,
+      "train_speed(iter/s)": 0.192291
+    },
+    {
+      "acc": 0.86829548,
+      "epoch": 0.5533465871438038,
+      "grad_norm": 8.956941093648082,
+      "learning_rate": 9.261426721804117e-06,
+      "loss": 0.65413742,
+      "memory(GiB)": 33.01,
+      "step": 1670,
+      "train_speed(iter/s)": 0.192295
+    },
+    {
+      "acc": 0.87451134,
+      "epoch": 0.5550033134526177,
+      "grad_norm": 3.616833903612867,
+      "learning_rate": 9.265157885548758e-06,
+      "loss": 0.64173145,
+      "memory(GiB)": 33.01,
+      "step": 1675,
+      "train_speed(iter/s)": 0.192298
+    },
+    {
+      "acc": 0.88274603,
+      "epoch": 0.5566600397614314,
+      "grad_norm": 6.213948043754148,
+      "learning_rate": 9.2688779280735e-06,
+      "loss": 0.62117977,
+      "memory(GiB)": 33.01,
+      "step": 1680,
+      "train_speed(iter/s)": 0.192301
+    },
+    {
+      "acc": 0.87925549,
+      "epoch": 0.5583167660702452,
+      "grad_norm": 4.50004747939087,
+      "learning_rate": 9.272586915477856e-06,
+      "loss": 0.63414607,
+      "memory(GiB)": 33.01,
+      "step": 1685,
+      "train_speed(iter/s)": 0.192304
+    },
+    {
+      "acc": 0.87159424,
+      "epoch": 0.559973492379059,
+      "grad_norm": 7.536326528887464,
+      "learning_rate": 9.276284913273801e-06,
+      "loss": 0.65830579,
+      "memory(GiB)": 33.01,
+      "step": 1690,
+      "train_speed(iter/s)": 0.192307
+    },
+    {
+      "acc": 0.87907772,
+      "epoch": 0.5616302186878728,
+      "grad_norm": 3.8732555461468636,
+      "learning_rate": 9.27997198639269e-06,
+      "loss": 0.63627729,
+      "memory(GiB)": 33.01,
+      "step": 1695,
+      "train_speed(iter/s)": 0.19231
+    },
+    {
+      "acc": 0.8568985,
+      "epoch": 0.5632869449966865,
+      "grad_norm": 3.6926446557268657,
+      "learning_rate": 9.283648199192115e-06,
+      "loss": 0.68832617,
+      "memory(GiB)": 33.01,
+      "step": 1700,
+      "train_speed(iter/s)": 0.192314
+    },
+    {
+      "acc": 0.86911259,
+      "epoch": 0.5649436713055004,
+      "grad_norm": 4.167878346282884,
+      "learning_rate": 9.287313615462637e-06,
+      "loss": 0.6848526,
+      "memory(GiB)": 33.01,
+      "step": 1705,
+      "train_speed(iter/s)": 0.192317
+    },
+    {
+      "acc": 0.8830162,
+      "epoch": 0.5666003976143141,
+      "grad_norm": 4.031298965260666,
+      "learning_rate": 9.290968298434428e-06,
+      "loss": 0.62251148,
+      "memory(GiB)": 33.01,
+      "step": 1710,
+      "train_speed(iter/s)": 0.192321
+    },
+    {
+      "acc": 0.87023563,
+      "epoch": 0.5682571239231279,
+      "grad_norm": 4.574398905309359,
+      "learning_rate": 9.294612310783819e-06,
+      "loss": 0.68600483,
+      "memory(GiB)": 33.01,
+      "step": 1715,
+      "train_speed(iter/s)": 0.192324
+    },
+    {
+      "acc": 0.86422176,
+      "epoch": 0.5699138502319416,
+      "grad_norm": 5.657723860144203,
+      "learning_rate": 9.298245714639748e-06,
+      "loss": 0.67713032,
+      "memory(GiB)": 33.01,
+      "step": 1720,
+      "train_speed(iter/s)": 0.192327
+    },
+    {
+      "acc": 0.85824757,
+      "epoch": 0.5715705765407555,
+      "grad_norm": 6.949231919339707,
+      "learning_rate": 9.301868571590115e-06,
+      "loss": 0.72136383,
+      "memory(GiB)": 33.01,
+      "step": 1725,
+      "train_speed(iter/s)": 0.192329
+    },
+    {
+      "acc": 0.86491108,
+      "epoch": 0.5732273028495692,
+      "grad_norm": 4.511148122728733,
+      "learning_rate": 9.305480942688047e-06,
+      "loss": 0.65395832,
+      "memory(GiB)": 33.01,
+      "step": 1730,
+      "train_speed(iter/s)": 0.192333
+    },
+    {
+      "acc": 0.85937595,
+      "epoch": 0.574884029158383,
+      "grad_norm": 6.076865498510463,
+      "learning_rate": 9.309082888458076e-06,
+      "loss": 0.67800088,
+      "memory(GiB)": 33.01,
+      "step": 1735,
+      "train_speed(iter/s)": 0.192337
+    },
+    {
+      "acc": 0.85908699,
+      "epoch": 0.5765407554671969,
+      "grad_norm": 6.640583907081719,
+      "learning_rate": 9.312674468902207e-06,
+      "loss": 0.68193717,
+      "memory(GiB)": 33.01,
+      "step": 1740,
+      "train_speed(iter/s)": 0.19234
+    },
+    {
+      "acc": 0.86825771,
+      "epoch": 0.5781974817760106,
+      "grad_norm": 2.9913874023324505,
+      "learning_rate": 9.316255743505935e-06,
+      "loss": 0.65605841,
+      "memory(GiB)": 33.01,
+      "step": 1745,
+      "train_speed(iter/s)": 0.192328
+    },
+    {
+      "acc": 0.86895323,
+      "epoch": 0.5798542080848244,
+      "grad_norm": 3.509690754353086,
+      "learning_rate": 9.319826771244152e-06,
+      "loss": 0.6424942,
+      "memory(GiB)": 33.01,
+      "step": 1750,
+      "train_speed(iter/s)": 0.192331
+    },
+    {
+      "acc": 0.87688522,
+      "epoch": 0.5815109343936382,
+      "grad_norm": 5.585633436797879,
+      "learning_rate": 9.32338761058696e-06,
+      "loss": 0.64732981,
+      "memory(GiB)": 33.01,
+      "step": 1755,
+      "train_speed(iter/s)": 0.192328
+    },
+    {
+      "acc": 0.86736622,
+      "epoch": 0.583167660702452,
+      "grad_norm": 12.021249469029797,
+      "learning_rate": 9.326938319505438e-06,
+      "loss": 0.69735794,
+      "memory(GiB)": 33.01,
+      "step": 1760,
+      "train_speed(iter/s)": 0.192332
+    },
+    {
+      "acc": 0.85666466,
+      "epoch": 0.5848243870112657,
+      "grad_norm": 9.141066601954984,
+      "learning_rate": 9.330478955477281e-06,
+      "loss": 0.76363716,
+      "memory(GiB)": 33.01,
+      "step": 1765,
+      "train_speed(iter/s)": 0.192335
+    },
+    {
+      "acc": 0.86701384,
+      "epoch": 0.5864811133200796,
+      "grad_norm": 5.115855433133116,
+      "learning_rate": 9.334009575492406e-06,
+      "loss": 0.69254465,
+      "memory(GiB)": 33.01,
+      "step": 1770,
+      "train_speed(iter/s)": 0.192339
+    },
+    {
+      "acc": 0.87474461,
+      "epoch": 0.5881378396288933,
+      "grad_norm": 2.613935884238985,
+      "learning_rate": 9.337530236058433e-06,
+      "loss": 0.61972847,
+      "memory(GiB)": 33.01,
+      "step": 1775,
+      "train_speed(iter/s)": 0.192343
+    },
+    {
+      "acc": 0.8837429,
+      "epoch": 0.5897945659377071,
+      "grad_norm": 2.765768204997492,
+      "learning_rate": 9.341040993206135e-06,
+      "loss": 0.59840755,
+      "memory(GiB)": 33.01,
+      "step": 1780,
+      "train_speed(iter/s)": 0.192346
+    },
+    {
+      "acc": 0.87598438,
+      "epoch": 0.5914512922465208,
+      "grad_norm": 3.902385923622911,
+      "learning_rate": 9.344541902494764e-06,
+      "loss": 0.63060737,
+      "memory(GiB)": 33.01,
+      "step": 1785,
+      "train_speed(iter/s)": 0.192349
+    },
+    {
+      "acc": 0.87122345,
+      "epoch": 0.5931080185553347,
+      "grad_norm": 8.347631641211056,
+      "learning_rate": 9.348033019017355e-06,
+      "loss": 0.63137841,
+      "memory(GiB)": 33.01,
+      "step": 1790,
+      "train_speed(iter/s)": 0.192353
+    },
+    {
+      "acc": 0.88335676,
+      "epoch": 0.5947647448641484,
+      "grad_norm": 7.931165061428975,
+      "learning_rate": 9.351514397405894e-06,
+      "loss": 0.61477599,
+      "memory(GiB)": 33.01,
+      "step": 1795,
+      "train_speed(iter/s)": 0.192356
+    },
+    {
+      "acc": 0.88620911,
+      "epoch": 0.5964214711729622,
+      "grad_norm": 6.496412648981423,
+      "learning_rate": 9.35498609183648e-06,
+      "loss": 0.59670343,
+      "memory(GiB)": 33.01,
+      "step": 1800,
+      "train_speed(iter/s)": 0.19236
+    },
+    {
+      "acc": 0.87989988,
+      "epoch": 0.598078197481776,
+      "grad_norm": 5.81429505167616,
+      "learning_rate": 9.358448156034365e-06,
+      "loss": 0.58589425,
+      "memory(GiB)": 33.01,
+      "step": 1805,
+      "train_speed(iter/s)": 0.192364
+    },
+    {
+      "acc": 0.86021442,
+      "epoch": 0.5997349237905898,
+      "grad_norm": 7.387526808154697,
+      "learning_rate": 9.361900643278946e-06,
+      "loss": 0.72263103,
+      "memory(GiB)": 33.01,
+      "step": 1810,
+      "train_speed(iter/s)": 0.192367
+    },
+    {
+      "acc": 0.8685957,
+      "epoch": 0.6013916500994035,
+      "grad_norm": 3.295357400629394,
+      "learning_rate": 9.365343606408687e-06,
+      "loss": 0.6860177,
+      "memory(GiB)": 33.01,
+      "step": 1815,
+      "train_speed(iter/s)": 0.192372
+    },
+    {
+      "acc": 0.87111511,
+      "epoch": 0.6030483764082174,
+      "grad_norm": 3.2613474978085786,
+      "learning_rate": 9.368777097825971e-06,
+      "loss": 0.65456285,
+      "memory(GiB)": 33.01,
+      "step": 1820,
+      "train_speed(iter/s)": 0.192377
+    },
+    {
+      "acc": 0.86717224,
+      "epoch": 0.6047051027170312,
+      "grad_norm": 3.6842014911472183,
+      "learning_rate": 9.37220116950188e-06,
+      "loss": 0.65306549,
+      "memory(GiB)": 33.01,
+      "step": 1825,
+      "train_speed(iter/s)": 0.19238
+    },
+    {
+      "acc": 0.86679745,
+      "epoch": 0.6063618290258449,
+      "grad_norm": 7.862766096911152,
+      "learning_rate": 9.375615872980926e-06,
+      "loss": 0.67689962,
+      "memory(GiB)": 33.01,
+      "step": 1830,
+      "train_speed(iter/s)": 0.192385
+    },
+    {
+      "acc": 0.85430756,
+      "epoch": 0.6080185553346588,
+      "grad_norm": 4.942428662598694,
+      "learning_rate": 9.379021259385697e-06,
+      "loss": 0.70608482,
+      "memory(GiB)": 33.01,
+      "step": 1835,
+      "train_speed(iter/s)": 0.19239
+    },
+    {
+      "acc": 0.87286358,
+      "epoch": 0.6096752816434725,
+      "grad_norm": 3.6884051952135573,
+      "learning_rate": 9.382417379421453e-06,
+      "loss": 0.65542011,
+      "memory(GiB)": 33.01,
+      "step": 1840,
+      "train_speed(iter/s)": 0.192397
+    },
+    {
+      "acc": 0.87217827,
+      "epoch": 0.6113320079522863,
+      "grad_norm": 4.46087116014083,
+      "learning_rate": 9.385804283380658e-06,
+      "loss": 0.63779249,
+      "memory(GiB)": 33.01,
+      "step": 1845,
+      "train_speed(iter/s)": 0.192402
+    },
+    {
+      "acc": 0.87114782,
+      "epoch": 0.6129887342611001,
+      "grad_norm": 8.164798448258196,
+      "learning_rate": 9.38918202114744e-06,
+      "loss": 0.65047865,
+      "memory(GiB)": 33.01,
+      "step": 1850,
+      "train_speed(iter/s)": 0.192408
+    },
+    {
+      "acc": 0.87397671,
+      "epoch": 0.6146454605699139,
+      "grad_norm": 8.406113413807082,
+      "learning_rate": 9.392550642202016e-06,
+      "loss": 0.62601881,
+      "memory(GiB)": 33.01,
+      "step": 1855,
+      "train_speed(iter/s)": 0.192412
+    },
+    {
+      "acc": 0.87091293,
+      "epoch": 0.6163021868787276,
+      "grad_norm": 4.973757562448072,
+      "learning_rate": 9.395910195625019e-06,
+      "loss": 0.63574677,
+      "memory(GiB)": 33.01,
+      "step": 1860,
+      "train_speed(iter/s)": 0.192417
+    },
+    {
+      "acc": 0.87237234,
+      "epoch": 0.6179589131875414,
+      "grad_norm": 7.259794102040732,
+      "learning_rate": 9.399260730101813e-06,
+      "loss": 0.65975394,
+      "memory(GiB)": 33.01,
+      "step": 1865,
+      "train_speed(iter/s)": 0.192421
+    },
+    {
+      "acc": 0.87357645,
+      "epoch": 0.6196156394963552,
+      "grad_norm": 4.948254368813279,
+      "learning_rate": 9.402602293926707e-06,
+      "loss": 0.66142483,
+      "memory(GiB)": 33.01,
+      "step": 1870,
+      "train_speed(iter/s)": 0.192425
+    },
+    {
+      "acc": 0.86736622,
+      "epoch": 0.621272365805169,
+      "grad_norm": 10.71133344628175,
+      "learning_rate": 9.405934935007133e-06,
+      "loss": 0.7003459,
+      "memory(GiB)": 33.01,
+      "step": 1875,
+      "train_speed(iter/s)": 0.192431
+    },
+    {
+      "acc": 0.86193829,
+      "epoch": 0.6229290921139827,
+      "grad_norm": 12.594851237762438,
+      "learning_rate": 9.40925870086779e-06,
+      "loss": 0.7385016,
+      "memory(GiB)": 33.01,
+      "step": 1880,
+      "train_speed(iter/s)": 0.192435
+    },
+    {
+      "acc": 0.8484436,
+      "epoch": 0.6245858184227966,
+      "grad_norm": 8.874937082337999,
+      "learning_rate": 9.412573638654678e-06,
+      "loss": 0.72295904,
+      "memory(GiB)": 33.01,
+      "step": 1885,
+      "train_speed(iter/s)": 0.19244
+    },
+    {
+      "acc": 0.86035633,
+      "epoch": 0.6262425447316103,
+      "grad_norm": 6.769670782418024,
+      "learning_rate": 9.415879795139132e-06,
+      "loss": 0.74117517,
+      "memory(GiB)": 33.01,
+      "step": 1890,
+      "train_speed(iter/s)": 0.192446
+    },
+    {
+      "acc": 0.8743494,
+      "epoch": 0.6278992710404241,
+      "grad_norm": 5.426482655852995,
+      "learning_rate": 9.419177216721772e-06,
+      "loss": 0.64620824,
+      "memory(GiB)": 33.01,
+      "step": 1895,
+      "train_speed(iter/s)": 0.192449
+    },
+    {
+      "acc": 0.86501122,
+      "epoch": 0.629555997349238,
+      "grad_norm": 4.701645724724149,
+      "learning_rate": 9.42246594943642e-06,
+      "loss": 0.6478301,
+      "memory(GiB)": 33.01,
+      "step": 1900,
+      "train_speed(iter/s)": 0.192455
+    },
+    {
+      "acc": 0.86878777,
+      "epoch": 0.6312127236580517,
+      "grad_norm": 4.529349528660305,
+      "learning_rate": 9.425746038953943e-06,
+      "loss": 0.71059952,
+      "memory(GiB)": 33.01,
+      "step": 1905,
+      "train_speed(iter/s)": 0.19246
+    },
+    {
+      "acc": 0.86721001,
+      "epoch": 0.6328694499668654,
+      "grad_norm": 4.017732753580181,
+      "learning_rate": 9.42901753058607e-06,
+      "loss": 0.66993275,
+      "memory(GiB)": 33.01,
+      "step": 1910,
+      "train_speed(iter/s)": 0.192465
+    },
+    {
+      "acc": 0.85818319,
+      "epoch": 0.6345261762756793,
+      "grad_norm": 5.189942543150934,
+      "learning_rate": 9.432280469289143e-06,
+      "loss": 0.73596964,
+      "memory(GiB)": 33.01,
+      "step": 1915,
+      "train_speed(iter/s)": 0.19247
+    },
+    {
+      "acc": 0.8732935,
+      "epoch": 0.6361829025844931,
+      "grad_norm": 4.370475133792059,
+      "learning_rate": 9.43553489966782e-06,
+      "loss": 0.68717313,
+      "memory(GiB)": 33.01,
+      "step": 1920,
+      "train_speed(iter/s)": 0.192475
+    },
+    {
+      "acc": 0.85696383,
+      "epoch": 0.6378396288933068,
+      "grad_norm": 7.911266975338863,
+      "learning_rate": 9.43878086597874e-06,
+      "loss": 0.7042037,
+      "memory(GiB)": 33.01,
+      "step": 1925,
+      "train_speed(iter/s)": 0.19248
+    },
+    {
+      "acc": 0.86053925,
+      "epoch": 0.6394963552021206,
+      "grad_norm": 4.885586074549585,
+      "learning_rate": 9.442018412134133e-06,
+      "loss": 0.68545146,
+      "memory(GiB)": 33.01,
+      "step": 1930,
+      "train_speed(iter/s)": 0.192485
+    },
+    {
+      "acc": 0.87085686,
+      "epoch": 0.6411530815109344,
+      "grad_norm": 3.1038574062542796,
+      "learning_rate": 9.44524758170538e-06,
+      "loss": 0.66258116,
+      "memory(GiB)": 33.01,
+      "step": 1935,
+      "train_speed(iter/s)": 0.192491
+    },
+    {
+      "acc": 0.86892052,
+      "epoch": 0.6428098078197482,
+      "grad_norm": 3.7218539198228733,
+      "learning_rate": 9.448468417926538e-06,
+      "loss": 0.65869718,
+      "memory(GiB)": 33.01,
+      "step": 1940,
+      "train_speed(iter/s)": 0.192496
+    },
+    {
+      "acc": 0.87207308,
+      "epoch": 0.6444665341285619,
+      "grad_norm": 7.02007077139689,
+      "learning_rate": 9.451680963697802e-06,
+      "loss": 0.63378029,
+      "memory(GiB)": 33.01,
+      "step": 1945,
+      "train_speed(iter/s)": 0.192501
+    },
+    {
+      "acc": 0.87004795,
+      "epoch": 0.6461232604373758,
+      "grad_norm": 4.748978576741996,
+      "learning_rate": 9.454885261588954e-06,
+      "loss": 0.65019732,
+      "memory(GiB)": 33.01,
+      "step": 1950,
+      "train_speed(iter/s)": 0.192506
+    },
+    {
+      "acc": 0.87050247,
+      "epoch": 0.6477799867461895,
+      "grad_norm": 9.090275346449877,
+      "learning_rate": 9.458081353842721e-06,
+      "loss": 0.68803792,
+      "memory(GiB)": 33.01,
+      "step": 1955,
+      "train_speed(iter/s)": 0.192511
+    },
+    {
+      "acc": 0.85250912,
+      "epoch": 0.6494367130550033,
+      "grad_norm": 6.585603979797727,
+      "learning_rate": 9.46126928237814e-06,
+      "loss": 0.74050913,
+      "memory(GiB)": 33.01,
+      "step": 1960,
+      "train_speed(iter/s)": 0.192516
+    },
+    {
+      "acc": 0.86775932,
+      "epoch": 0.6510934393638171,
+      "grad_norm": 5.721359622127249,
+      "learning_rate": 9.464449088793843e-06,
+      "loss": 0.68673601,
+      "memory(GiB)": 33.01,
+      "step": 1965,
+      "train_speed(iter/s)": 0.192521
+    },
+    {
+      "acc": 0.85250912,
+      "epoch": 0.6527501656726309,
+      "grad_norm": 5.38426969316766,
+      "learning_rate": 9.467620814371318e-06,
+      "loss": 0.7278306,
+      "memory(GiB)": 33.01,
+      "step": 1970,
+      "train_speed(iter/s)": 0.192527
+    },
+    {
+      "acc": 0.87017155,
+      "epoch": 0.6544068919814446,
+      "grad_norm": 4.3969246956399735,
+      "learning_rate": 9.470784500078125e-06,
+      "loss": 0.69506721,
+      "memory(GiB)": 33.01,
+      "step": 1975,
+      "train_speed(iter/s)": 0.192516
+    },
+    {
+      "acc": 0.8760396,
+      "epoch": 0.6560636182902585,
+      "grad_norm": 2.313031591146396,
+      "learning_rate": 9.47394018657107e-06,
+      "loss": 0.63808465,
+      "memory(GiB)": 33.01,
+      "step": 1980,
+      "train_speed(iter/s)": 0.192516
+    },
+    {
+      "acc": 0.86906242,
+      "epoch": 0.6577203445990722,
+      "grad_norm": 3.4070078308033827,
+      "learning_rate": 9.477087914199344e-06,
+      "loss": 0.62466335,
+      "memory(GiB)": 33.01,
+      "step": 1985,
+      "train_speed(iter/s)": 0.192521
+    },
+    {
+      "acc": 0.87359676,
+      "epoch": 0.659377070907886,
+      "grad_norm": 5.308161945186524,
+      "learning_rate": 9.48022772300761e-06,
+      "loss": 0.67417727,
+      "memory(GiB)": 33.01,
+      "step": 1990,
+      "train_speed(iter/s)": 0.192528
+    },
+    {
+      "acc": 0.87883778,
+      "epoch": 0.6610337972166997,
+      "grad_norm": 5.225205276711047,
+      "learning_rate": 9.48335965273907e-06,
+      "loss": 0.63469763,
+      "memory(GiB)": 33.01,
+      "step": 1995,
+      "train_speed(iter/s)": 0.192533
+    },
+    {
+      "acc": 0.87891026,
+      "epoch": 0.6626905235255136,
+      "grad_norm": 5.226052280288049,
+      "learning_rate": 9.486483742838472e-06,
+      "loss": 0.63047624,
+      "memory(GiB)": 33.01,
+      "step": 2000,
+      "train_speed(iter/s)": 0.192539
+    },
+    {
+      "acc": 0.88026047,
+      "epoch": 0.6643472498343274,
+      "grad_norm": 8.743298332087017,
+      "learning_rate": 9.489600032455106e-06,
+      "loss": 0.61615877,
+      "memory(GiB)": 33.01,
+      "step": 2005,
+      "train_speed(iter/s)": 0.192545
+    },
+    {
+      "acc": 0.871735,
+      "epoch": 0.6660039761431411,
+      "grad_norm": 6.115424883254176,
+      "learning_rate": 9.492708560445731e-06,
+      "loss": 0.6272666,
+      "memory(GiB)": 33.01,
+      "step": 2010,
+      "train_speed(iter/s)": 0.19255
+    },
+    {
+      "acc": 0.86677151,
+      "epoch": 0.667660702451955,
+      "grad_norm": 5.472680938783484,
+      "learning_rate": 9.49580936537749e-06,
+      "loss": 0.69325242,
+      "memory(GiB)": 33.01,
+      "step": 2015,
+      "train_speed(iter/s)": 0.192555
+    },
+    {
+      "acc": 0.87383423,
+      "epoch": 0.6693174287607687,
+      "grad_norm": 7.415846100116536,
+      "learning_rate": 9.498902485530788e-06,
+      "loss": 0.64826474,
+      "memory(GiB)": 33.01,
+      "step": 2020,
+      "train_speed(iter/s)": 0.19256
+    },
+    {
+      "acc": 0.87591295,
+      "epoch": 0.6709741550695825,
+      "grad_norm": 4.961612622529448,
+      "learning_rate": 9.501987958902114e-06,
+      "loss": 0.64979839,
+      "memory(GiB)": 33.01,
+      "step": 2025,
+      "train_speed(iter/s)": 0.192565
+    },
+    {
+      "acc": 0.8747304,
+      "epoch": 0.6726308813783963,
+      "grad_norm": 3.7629995987160263,
+      "learning_rate": 9.505065823206848e-06,
+      "loss": 0.64468951,
+      "memory(GiB)": 33.01,
+      "step": 2030,
+      "train_speed(iter/s)": 0.19257
+    },
+    {
+      "acc": 0.8648468,
+      "epoch": 0.6742876076872101,
+      "grad_norm": 9.481191593906079,
+      "learning_rate": 9.50813611588203e-06,
+      "loss": 0.70373602,
+      "memory(GiB)": 33.01,
+      "step": 2035,
+      "train_speed(iter/s)": 0.192575
+    },
+    {
+      "acc": 0.87640219,
+      "epoch": 0.6759443339960238,
+      "grad_norm": 5.609814410215989,
+      "learning_rate": 9.511198874089086e-06,
+      "loss": 0.64934058,
+      "memory(GiB)": 33.01,
+      "step": 2040,
+      "train_speed(iter/s)": 0.192581
+    },
+    {
+      "acc": 0.85914211,
+      "epoch": 0.6776010603048377,
+      "grad_norm": 6.279392103025398,
+      "learning_rate": 9.514254134716537e-06,
+      "loss": 0.72303848,
+      "memory(GiB)": 33.01,
+      "step": 2045,
+      "train_speed(iter/s)": 0.192587
+    },
+    {
+      "acc": 0.86438618,
+      "epoch": 0.6792577866136514,
+      "grad_norm": 3.57928961851651,
+      "learning_rate": 9.51730193438265e-06,
+      "loss": 0.65385365,
+      "memory(GiB)": 33.01,
+      "step": 2050,
+      "train_speed(iter/s)": 0.192592
+    },
+    {
+      "acc": 0.86672287,
+      "epoch": 0.6809145129224652,
+      "grad_norm": 4.72146459698859,
+      "learning_rate": 9.52034230943808e-06,
+      "loss": 0.6649436,
+      "memory(GiB)": 33.01,
+      "step": 2055,
+      "train_speed(iter/s)": 0.192598
+    },
+    {
+      "acc": 0.86496525,
+      "epoch": 0.682571239231279,
+      "grad_norm": 3.9579348297855312,
+      "learning_rate": 9.52337529596848e-06,
+      "loss": 0.63084106,
+      "memory(GiB)": 33.01,
+      "step": 2060,
+      "train_speed(iter/s)": 0.192602
+    },
+    {
+      "acc": 0.85933104,
+      "epoch": 0.6842279655400928,
+      "grad_norm": 2.9487788456501884,
+      "learning_rate": 9.526400929797046e-06,
+      "loss": 0.69089756,
+      "memory(GiB)": 33.01,
+      "step": 2065,
+      "train_speed(iter/s)": 0.192609
+    },
+    {
+      "acc": 0.87197399,
+      "epoch": 0.6858846918489065,
+      "grad_norm": 6.630580178404849,
+      "learning_rate": 9.529419246487087e-06,
+      "loss": 0.68466773,
+      "memory(GiB)": 33.01,
+      "step": 2070,
+      "train_speed(iter/s)": 0.192614
+    },
+    {
+      "acc": 0.87497139,
+      "epoch": 0.6875414181577203,
+      "grad_norm": 5.261229659377707,
+      "learning_rate": 9.532430281344505e-06,
+      "loss": 0.65252924,
+      "memory(GiB)": 33.01,
+      "step": 2075,
+      "train_speed(iter/s)": 0.192619
+    },
+    {
+      "acc": 0.8781332,
+      "epoch": 0.6891981444665342,
+      "grad_norm": 5.791336126885148,
+      "learning_rate": 9.535434069420291e-06,
+      "loss": 0.64834123,
+      "memory(GiB)": 33.01,
+      "step": 2080,
+      "train_speed(iter/s)": 0.192624
+    },
+    {
+      "acc": 0.87972021,
+      "epoch": 0.6908548707753479,
+      "grad_norm": 6.244725426282354,
+      "learning_rate": 9.53843064551297e-06,
+      "loss": 0.60382318,
+      "memory(GiB)": 33.01,
+      "step": 2085,
+      "train_speed(iter/s)": 0.192629
+    },
+    {
+      "acc": 0.8692606,
+      "epoch": 0.6925115970841617,
+      "grad_norm": 3.3223224218287637,
+      "learning_rate": 9.54142004417101e-06,
+      "loss": 0.62219219,
+      "memory(GiB)": 33.01,
+      "step": 2090,
+      "train_speed(iter/s)": 0.192633
+    },
+    {
+      "acc": 0.86586094,
+      "epoch": 0.6941683233929755,
+      "grad_norm": 8.333006833340361,
+      "learning_rate": 9.54440229969522e-06,
+      "loss": 0.66129503,
+      "memory(GiB)": 33.01,
+      "step": 2095,
+      "train_speed(iter/s)": 0.192638
+    },
+    {
+      "acc": 0.87937078,
+      "epoch": 0.6958250497017893,
+      "grad_norm": 4.848522391848659,
+      "learning_rate": 9.547377446141123e-06,
+      "loss": 0.59719243,
+      "memory(GiB)": 33.01,
+      "step": 2100,
+      "train_speed(iter/s)": 0.192644
+    },
+    {
+      "acc": 0.87653017,
+      "epoch": 0.697481776010603,
+      "grad_norm": 6.058172300357878,
+      "learning_rate": 9.55034551732126e-06,
+      "loss": 0.6292902,
+      "memory(GiB)": 33.01,
+      "step": 2105,
+      "train_speed(iter/s)": 0.19265
+    },
+    {
+      "acc": 0.86440668,
+      "epoch": 0.6991385023194169,
+      "grad_norm": 5.422445091075002,
+      "learning_rate": 9.553306546807525e-06,
+      "loss": 0.65456238,
+      "memory(GiB)": 33.01,
+      "step": 2110,
+      "train_speed(iter/s)": 0.192656
+    },
+    {
+      "acc": 0.87416859,
+      "epoch": 0.7007952286282306,
+      "grad_norm": 5.593253876187401,
+      "learning_rate": 9.556260567933424e-06,
+      "loss": 0.63578682,
+      "memory(GiB)": 33.01,
+      "step": 2115,
+      "train_speed(iter/s)": 0.192661
+    },
+    {
+      "acc": 0.87171974,
+      "epoch": 0.7024519549370444,
+      "grad_norm": 5.7664175346274575,
+      "learning_rate": 9.559207613796336e-06,
+      "loss": 0.70321331,
+      "memory(GiB)": 33.01,
+      "step": 2120,
+      "train_speed(iter/s)": 0.192666
+    },
+    {
+      "acc": 0.8626276,
+      "epoch": 0.7041086812458582,
+      "grad_norm": 4.1724329783244345,
+      "learning_rate": 9.56214771725974e-06,
+      "loss": 0.65507898,
+      "memory(GiB)": 33.01,
+      "step": 2125,
+      "train_speed(iter/s)": 0.192671
+    },
+    {
+      "acc": 0.87635355,
+      "epoch": 0.705765407554672,
+      "grad_norm": 2.212910015605007,
+      "learning_rate": 9.565080910955405e-06,
+      "loss": 0.62007113,
+      "memory(GiB)": 33.01,
+      "step": 2130,
+      "train_speed(iter/s)": 0.192676
+    },
+    {
+      "acc": 0.8778676,
+      "epoch": 0.7074221338634857,
+      "grad_norm": 4.229729013305469,
+      "learning_rate": 9.568007227285568e-06,
+      "loss": 0.62818556,
+      "memory(GiB)": 33.01,
+      "step": 2135,
+      "train_speed(iter/s)": 0.192681
+    },
+    {
+      "acc": 0.86363354,
+      "epoch": 0.7090788601722995,
+      "grad_norm": 5.191226429441517,
+      "learning_rate": 9.570926698425084e-06,
+      "loss": 0.6825675,
+      "memory(GiB)": 33.01,
+      "step": 2140,
+      "train_speed(iter/s)": 0.192685
+    },
+    {
+      "acc": 0.87175541,
+      "epoch": 0.7107355864811133,
+      "grad_norm": 6.668956366113271,
+      "learning_rate": 9.573839356323542e-06,
+      "loss": 0.67435064,
+      "memory(GiB)": 33.01,
+      "step": 2145,
+      "train_speed(iter/s)": 0.192689
+    },
+    {
+      "acc": 0.86686783,
+      "epoch": 0.7123923127899271,
+      "grad_norm": 5.313814404839771,
+      "learning_rate": 9.576745232707372e-06,
+      "loss": 0.68711166,
+      "memory(GiB)": 33.01,
+      "step": 2150,
+      "train_speed(iter/s)": 0.192694
+    },
+    {
+      "acc": 0.8734129,
+      "epoch": 0.7140490390987408,
+      "grad_norm": 6.254802303643827,
+      "learning_rate": 9.57964435908191e-06,
+      "loss": 0.64754696,
+      "memory(GiB)": 33.01,
+      "step": 2155,
+      "train_speed(iter/s)": 0.192699
+    },
+    {
+      "acc": 0.86314135,
+      "epoch": 0.7157057654075547,
+      "grad_norm": 3.5194625347786626,
+      "learning_rate": 9.582536766733452e-06,
+      "loss": 0.68779392,
+      "memory(GiB)": 33.01,
+      "step": 2160,
+      "train_speed(iter/s)": 0.192705
+    },
+    {
+      "acc": 0.86745138,
+      "epoch": 0.7173624917163685,
+      "grad_norm": 7.667382579416203,
+      "learning_rate": 9.585422486731281e-06,
+      "loss": 0.64779177,
+      "memory(GiB)": 33.01,
+      "step": 2165,
+      "train_speed(iter/s)": 0.19271
+    },
+    {
+      "acc": 0.85620298,
+      "epoch": 0.7190192180251822,
+      "grad_norm": 11.09618897417527,
+      "learning_rate": 9.588301549929662e-06,
+      "loss": 0.79766526,
+      "memory(GiB)": 33.01,
+      "step": 2170,
+      "train_speed(iter/s)": 0.192714
+    },
+    {
+      "acc": 0.88470783,
+      "epoch": 0.7206759443339961,
+      "grad_norm": 4.853353795149172,
+      "learning_rate": 9.59117398696983e-06,
+      "loss": 0.66207132,
+      "memory(GiB)": 33.01,
+      "step": 2175,
+      "train_speed(iter/s)": 0.19272
+    },
+    {
+      "acc": 0.8712225,
+      "epoch": 0.7223326706428098,
+      "grad_norm": 4.423883047425217,
+      "learning_rate": 9.594039828281947e-06,
+      "loss": 0.63862867,
+      "memory(GiB)": 33.01,
+      "step": 2180,
+      "train_speed(iter/s)": 0.192725
+    },
+    {
+      "acc": 0.87145014,
+      "epoch": 0.7239893969516236,
+      "grad_norm": 3.296421168238937,
+      "learning_rate": 9.596899104087026e-06,
+      "loss": 0.65805697,
+      "memory(GiB)": 33.01,
+      "step": 2185,
+      "train_speed(iter/s)": 0.19273
+    },
+    {
+      "acc": 0.8778677,
+      "epoch": 0.7256461232604374,
+      "grad_norm": 5.366937991553024,
+      "learning_rate": 9.599751844398853e-06,
+      "loss": 0.64201961,
+      "memory(GiB)": 33.01,
+      "step": 2190,
+      "train_speed(iter/s)": 0.192735
+    },
+    {
+      "acc": 0.8649683,
+      "epoch": 0.7273028495692512,
+      "grad_norm": 4.97366153444764,
+      "learning_rate": 9.602598079025871e-06,
+      "loss": 0.66631365,
+      "memory(GiB)": 33.01,
+      "step": 2195,
+      "train_speed(iter/s)": 0.192739
+    },
+    {
+      "acc": 0.86420956,
+      "epoch": 0.7289595758780649,
+      "grad_norm": 4.878490585708368,
+      "learning_rate": 9.605437837573062e-06,
+      "loss": 0.66615968,
+      "memory(GiB)": 33.01,
+      "step": 2200,
+      "train_speed(iter/s)": 0.192739
+    },
+    {
+      "acc": 0.8645731,
+      "epoch": 0.7306163021868787,
+      "grad_norm": 9.691422276593839,
+      "learning_rate": 9.608271149443773e-06,
+      "loss": 0.66423512,
+      "memory(GiB)": 33.01,
+      "step": 2205,
+      "train_speed(iter/s)": 0.19273
+    },
+    {
+      "acc": 0.8680954,
+      "epoch": 0.7322730284956925,
+      "grad_norm": 5.243986361705653,
+      "learning_rate": 9.61109804384156e-06,
+      "loss": 0.68853683,
+      "memory(GiB)": 33.01,
+      "step": 2210,
+      "train_speed(iter/s)": 0.19273
+    },
+    {
+      "acc": 0.87544937,
+      "epoch": 0.7339297548045063,
+      "grad_norm": 4.484903741087824,
+      "learning_rate": 9.613918549771985e-06,
+      "loss": 0.64813089,
+      "memory(GiB)": 33.01,
+      "step": 2215,
+      "train_speed(iter/s)": 0.192736
+    },
+    {
+      "acc": 0.87906456,
+      "epoch": 0.73558648111332,
+      "grad_norm": 3.290790298652583,
+      "learning_rate": 9.616732696044411e-06,
+      "loss": 0.61578646,
+      "memory(GiB)": 33.01,
+      "step": 2220,
+      "train_speed(iter/s)": 0.19274
+    },
+    {
+      "acc": 0.8714159,
+      "epoch": 0.7372432074221339,
+      "grad_norm": 4.544670383323588,
+      "learning_rate": 9.619540511273759e-06,
+      "loss": 0.66664362,
+      "memory(GiB)": 33.01,
+      "step": 2225,
+      "train_speed(iter/s)": 0.192745
+    },
+    {
+      "acc": 0.87207813,
+      "epoch": 0.7388999337309476,
+      "grad_norm": 4.6854579440589275,
+      "learning_rate": 9.622342023882252e-06,
+      "loss": 0.66553741,
+      "memory(GiB)": 33.01,
+      "step": 2230,
+      "train_speed(iter/s)": 0.192751
+    },
+    {
+      "acc": 0.86470995,
+      "epoch": 0.7405566600397614,
+      "grad_norm": 5.073985532829242,
+      "learning_rate": 9.625137262101154e-06,
+      "loss": 0.68595324,
+      "memory(GiB)": 33.01,
+      "step": 2235,
+      "train_speed(iter/s)": 0.192755
+    },
+    {
+      "acc": 0.87021656,
+      "epoch": 0.7422133863485753,
+      "grad_norm": 3.4425971798138897,
+      "learning_rate": 9.627926253972462e-06,
+      "loss": 0.66654301,
+      "memory(GiB)": 33.01,
+      "step": 2240,
+      "train_speed(iter/s)": 0.19276
+    },
+    {
+      "acc": 0.86235189,
+      "epoch": 0.743870112657389,
+      "grad_norm": 3.8965782786196916,
+      "learning_rate": 9.630709027350606e-06,
+      "loss": 0.66385617,
+      "memory(GiB)": 33.01,
+      "step": 2245,
+      "train_speed(iter/s)": 0.192765
+    },
+    {
+      "acc": 0.86006021,
+      "epoch": 0.7455268389662028,
+      "grad_norm": 5.99644345544046,
+      "learning_rate": 9.633485609904106e-06,
+      "loss": 0.69552994,
+      "memory(GiB)": 33.01,
+      "step": 2250,
+      "train_speed(iter/s)": 0.192769
+    },
+    {
+      "acc": 0.87564688,
+      "epoch": 0.7471835652750166,
+      "grad_norm": 2.8001309984832328,
+      "learning_rate": 9.63625602911724e-06,
+      "loss": 0.63839378,
+      "memory(GiB)": 33.01,
+      "step": 2255,
+      "train_speed(iter/s)": 0.192773
+    },
+    {
+      "acc": 0.87822104,
+      "epoch": 0.7488402915838304,
+      "grad_norm": 4.297893306015321,
+      "learning_rate": 9.639020312291654e-06,
+      "loss": 0.61689587,
+      "memory(GiB)": 33.01,
+      "step": 2260,
+      "train_speed(iter/s)": 0.192779
+    },
+    {
+      "acc": 0.86783085,
+      "epoch": 0.7504970178926441,
+      "grad_norm": 26.487082811616887,
+      "learning_rate": 9.641778486548e-06,
+      "loss": 0.66026106,
+      "memory(GiB)": 33.01,
+      "step": 2265,
+      "train_speed(iter/s)": 0.192783
+    },
+    {
+      "acc": 0.87071695,
+      "epoch": 0.752153744201458,
+      "grad_norm": 3.2521989867655363,
+      "learning_rate": 9.644530578827511e-06,
+      "loss": 0.63276267,
+      "memory(GiB)": 33.01,
+      "step": 2270,
+      "train_speed(iter/s)": 0.192789
+    },
+    {
+      "acc": 0.86867542,
+      "epoch": 0.7538104705102717,
+      "grad_norm": 3.8223108119868083,
+      "learning_rate": 9.647276615893595e-06,
+      "loss": 0.68160901,
+      "memory(GiB)": 33.01,
+      "step": 2275,
+      "train_speed(iter/s)": 0.192793
+    },
+    {
+      "acc": 0.87834759,
+      "epoch": 0.7554671968190855,
+      "grad_norm": 3.939928542285973,
+      "learning_rate": 9.65001662433339e-06,
+      "loss": 0.62497997,
+      "memory(GiB)": 33.01,
+      "step": 2280,
+      "train_speed(iter/s)": 0.192798
+    },
+    {
+      "acc": 0.87994862,
+      "epoch": 0.7571239231278992,
+      "grad_norm": 3.6752016267248178,
+      "learning_rate": 9.652750630559314e-06,
+      "loss": 0.60249023,
+      "memory(GiB)": 33.01,
+      "step": 2285,
+      "train_speed(iter/s)": 0.192803
+    },
+    {
+      "acc": 0.86858349,
+      "epoch": 0.7587806494367131,
+      "grad_norm": 6.696075785561995,
+      "learning_rate": 9.655478660810578e-06,
+      "loss": 0.69671068,
+      "memory(GiB)": 33.01,
+      "step": 2290,
+      "train_speed(iter/s)": 0.192809
+    },
+    {
+      "acc": 0.86544838,
+      "epoch": 0.7604373757455268,
+      "grad_norm": 4.977378905291485,
+      "learning_rate": 9.65820074115472e-06,
+      "loss": 0.68020229,
+      "memory(GiB)": 33.01,
+      "step": 2295,
+      "train_speed(iter/s)": 0.192814
+    },
+    {
+      "acc": 0.87358656,
+      "epoch": 0.7620941020543406,
+      "grad_norm": 8.906192831262254,
+      "learning_rate": 9.660916897489078e-06,
+      "loss": 0.64444475,
+      "memory(GiB)": 33.01,
+      "step": 2300,
+      "train_speed(iter/s)": 0.192818
+    },
+    {
+      "acc": 0.87565355,
+      "epoch": 0.7637508283631544,
+      "grad_norm": 3.68508372434642,
+      "learning_rate": 9.663627155542284e-06,
+      "loss": 0.66665568,
+      "memory(GiB)": 33.01,
+      "step": 2305,
+      "train_speed(iter/s)": 0.192823
+    },
+    {
+      "acc": 0.87511635,
+      "epoch": 0.7654075546719682,
+      "grad_norm": 4.518392370164668,
+      "learning_rate": 9.666331540875712e-06,
+      "loss": 0.63670602,
+      "memory(GiB)": 33.01,
+      "step": 2310,
+      "train_speed(iter/s)": 0.192828
+    },
+    {
+      "acc": 0.87756433,
+      "epoch": 0.7670642809807819,
+      "grad_norm": 3.189881236903923,
+      "learning_rate": 9.66903007888494e-06,
+      "loss": 0.59702106,
+      "memory(GiB)": 33.01,
+      "step": 2315,
+      "train_speed(iter/s)": 0.192833
+    },
+    {
+      "acc": 0.87794418,
+      "epoch": 0.7687210072895958,
+      "grad_norm": 6.921943736323694,
+      "learning_rate": 9.67172279480117e-06,
+      "loss": 0.59983635,
+      "memory(GiB)": 33.01,
+      "step": 2320,
+      "train_speed(iter/s)": 0.192837
+    },
+    {
+      "acc": 0.87574348,
+      "epoch": 0.7703777335984096,
+      "grad_norm": 4.915893162263451,
+      "learning_rate": 9.674409713692644e-06,
+      "loss": 0.60740991,
+      "memory(GiB)": 33.01,
+      "step": 2325,
+      "train_speed(iter/s)": 0.192842
+    },
+    {
+      "acc": 0.87510929,
+      "epoch": 0.7720344599072233,
+      "grad_norm": 11.334691577781541,
+      "learning_rate": 9.677090860466051e-06,
+      "loss": 0.63948035,
+      "memory(GiB)": 33.01,
+      "step": 2330,
+      "train_speed(iter/s)": 0.192846
+    },
+    {
+      "acc": 0.8721446,
+      "epoch": 0.7736911862160372,
+      "grad_norm": 6.617023449279387,
+      "learning_rate": 9.679766259867908e-06,
+      "loss": 0.71691961,
+      "memory(GiB)": 33.01,
+      "step": 2335,
+      "train_speed(iter/s)": 0.19285
+    },
+    {
+      "acc": 0.87824421,
+      "epoch": 0.7753479125248509,
+      "grad_norm": 5.3101785187888195,
+      "learning_rate": 9.682435936485925e-06,
+      "loss": 0.62746267,
+      "memory(GiB)": 33.01,
+      "step": 2340,
+      "train_speed(iter/s)": 0.192855
+    },
+    {
+      "acc": 0.87435551,
+      "epoch": 0.7770046388336647,
+      "grad_norm": 6.644924418489874,
+      "learning_rate": 9.685099914750372e-06,
+      "loss": 0.62730703,
+      "memory(GiB)": 33.01,
+      "step": 2345,
+      "train_speed(iter/s)": 0.192859
+    },
+    {
+      "acc": 0.87127953,
+      "epoch": 0.7786613651424784,
+      "grad_norm": 3.5375242584252793,
+      "learning_rate": 9.687758218935415e-06,
+      "loss": 0.6645195,
+      "memory(GiB)": 33.01,
+      "step": 2350,
+      "train_speed(iter/s)": 0.192863
+    },
+    {
+      "acc": 0.87114382,
+      "epoch": 0.7803180914512923,
+      "grad_norm": 5.52235553077396,
+      "learning_rate": 9.69041087316044e-06,
+      "loss": 0.63387938,
+      "memory(GiB)": 33.01,
+      "step": 2355,
+      "train_speed(iter/s)": 0.192868
+    },
+    {
+      "acc": 0.87978954,
+      "epoch": 0.781974817760106,
+      "grad_norm": 6.25879731840057,
+      "learning_rate": 9.693057901391369e-06,
+      "loss": 0.62484179,
+      "memory(GiB)": 33.01,
+      "step": 2360,
+      "train_speed(iter/s)": 0.192873
+    },
+    {
+      "acc": 0.87951393,
+      "epoch": 0.7836315440689198,
+      "grad_norm": 5.480847419045661,
+      "learning_rate": 9.695699327441961e-06,
+      "loss": 0.63349662,
+      "memory(GiB)": 33.01,
+      "step": 2365,
+      "train_speed(iter/s)": 0.192879
+    },
+    {
+      "acc": 0.86821175,
+      "epoch": 0.7852882703777336,
+      "grad_norm": 5.457799408858417,
+      "learning_rate": 9.698335174975095e-06,
+      "loss": 0.64089489,
+      "memory(GiB)": 33.01,
+      "step": 2370,
+      "train_speed(iter/s)": 0.192883
+    },
+    {
+      "acc": 0.8801096,
+      "epoch": 0.7869449966865474,
+      "grad_norm": 3.4325244782448894,
+      "learning_rate": 9.700965467504045e-06,
+      "loss": 0.61814466,
+      "memory(GiB)": 33.01,
+      "step": 2375,
+      "train_speed(iter/s)": 0.192887
+    },
+    {
+      "acc": 0.87665138,
+      "epoch": 0.7886017229953611,
+      "grad_norm": 6.683363108167596,
+      "learning_rate": 9.703590228393729e-06,
+      "loss": 0.59398308,
+      "memory(GiB)": 33.01,
+      "step": 2380,
+      "train_speed(iter/s)": 0.192891
+    },
+    {
+      "acc": 0.87479162,
+      "epoch": 0.790258449304175,
+      "grad_norm": 3.3726614620556146,
+      "learning_rate": 9.70620948086197e-06,
+      "loss": 0.6071887,
+      "memory(GiB)": 33.01,
+      "step": 2385,
+      "train_speed(iter/s)": 0.192895
+    },
+    {
+      "acc": 0.87130003,
+      "epoch": 0.7919151756129887,
+      "grad_norm": 7.113120412038941,
+      "learning_rate": 9.708823247980712e-06,
+      "loss": 0.6394248,
+      "memory(GiB)": 33.01,
+      "step": 2390,
+      "train_speed(iter/s)": 0.192899
+    },
+    {
+      "acc": 0.85982437,
+      "epoch": 0.7935719019218025,
+      "grad_norm": 10.05497132850973,
+      "learning_rate": 9.711431552677255e-06,
+      "loss": 0.67967396,
+      "memory(GiB)": 33.01,
+      "step": 2395,
+      "train_speed(iter/s)": 0.192903
+    },
+    {
+      "acc": 0.89040947,
+      "epoch": 0.7952286282306164,
+      "grad_norm": 4.460223038735573,
+      "learning_rate": 9.714034417735445e-06,
+      "loss": 0.58384981,
+      "memory(GiB)": 33.01,
+      "step": 2400,
+      "train_speed(iter/s)": 0.192908
+    },
+    {
+      "acc": 0.87658396,
+      "epoch": 0.7968853545394301,
+      "grad_norm": 2.9971847928295734,
+      "learning_rate": 9.716631865796885e-06,
+      "loss": 0.59908514,
+      "memory(GiB)": 33.01,
+      "step": 2405,
+      "train_speed(iter/s)": 0.192911
+    },
+    {
+      "acc": 0.88406553,
+      "epoch": 0.7985420808482439,
+      "grad_norm": 5.596120844203542,
+      "learning_rate": 9.719223919362102e-06,
+      "loss": 0.59015636,
+      "memory(GiB)": 33.01,
+      "step": 2410,
+      "train_speed(iter/s)": 0.192916
+    },
+    {
+      "acc": 0.87028484,
+      "epoch": 0.8001988071570576,
+      "grad_norm": 7.31459559392489,
+      "learning_rate": 9.721810600791728e-06,
+      "loss": 0.66939793,
+      "memory(GiB)": 33.01,
+      "step": 2415,
+      "train_speed(iter/s)": 0.19292
+    },
+    {
+      "acc": 0.88224373,
+      "epoch": 0.8018555334658715,
+      "grad_norm": 5.106041137712201,
+      "learning_rate": 9.72439193230765e-06,
+      "loss": 0.58553267,
+      "memory(GiB)": 33.01,
+      "step": 2420,
+      "train_speed(iter/s)": 0.192924
+    },
+    {
+      "acc": 0.88108397,
+      "epoch": 0.8035122597746852,
+      "grad_norm": 3.6988250041450312,
+      "learning_rate": 9.726967935994161e-06,
+      "loss": 0.66949582,
+      "memory(GiB)": 33.01,
+      "step": 2425,
+      "train_speed(iter/s)": 0.192928
+    },
+    {
+      "acc": 0.86570663,
+      "epoch": 0.805168986083499,
+      "grad_norm": 9.916491784398852,
+      "learning_rate": 9.729538633799085e-06,
+      "loss": 0.66917977,
+      "memory(GiB)": 33.01,
+      "step": 2430,
+      "train_speed(iter/s)": 0.19292
+    },
+    {
+      "acc": 0.86554222,
+      "epoch": 0.8068257123923128,
+      "grad_norm": 9.44945007567293,
+      "learning_rate": 9.73210404753491e-06,
+      "loss": 0.68716011,
+      "memory(GiB)": 33.01,
+      "step": 2435,
+      "train_speed(iter/s)": 0.192915
+    },
+    {
+      "acc": 0.86351709,
+      "epoch": 0.8084824387011266,
+      "grad_norm": 6.906873336229187,
+      "learning_rate": 9.73466419887989e-06,
+      "loss": 0.7135231,
+      "memory(GiB)": 33.01,
+      "step": 2440,
+      "train_speed(iter/s)": 0.19292
+    },
+    {
+      "acc": 0.86703777,
+      "epoch": 0.8101391650099403,
+      "grad_norm": 7.012954588113596,
+      "learning_rate": 9.73721910937915e-06,
+      "loss": 0.68227029,
+      "memory(GiB)": 33.01,
+      "step": 2445,
+      "train_speed(iter/s)": 0.192924
+    },
+    {
+      "acc": 0.87811565,
+      "epoch": 0.8117958913187542,
+      "grad_norm": 4.698961796410276,
+      "learning_rate": 9.739768800445764e-06,
+      "loss": 0.63987741,
+      "memory(GiB)": 33.01,
+      "step": 2450,
+      "train_speed(iter/s)": 0.192928
+    },
+    {
+      "acc": 0.86939335,
+      "epoch": 0.8134526176275679,
+      "grad_norm": 4.205330590419783,
+      "learning_rate": 9.742313293361852e-06,
+      "loss": 0.621698,
+      "memory(GiB)": 33.01,
+      "step": 2455,
+      "train_speed(iter/s)": 0.192932
+    },
+    {
+      "acc": 0.86712561,
+      "epoch": 0.8151093439363817,
+      "grad_norm": 4.496722035371473,
+      "learning_rate": 9.74485260927962e-06,
+      "loss": 0.65557423,
+      "memory(GiB)": 33.01,
+      "step": 2460,
+      "train_speed(iter/s)": 0.192937
+    },
+    {
+      "acc": 0.89136581,
+      "epoch": 0.8167660702451955,
+      "grad_norm": 5.541308303048364,
+      "learning_rate": 9.747386769222437e-06,
+      "loss": 0.59083924,
+      "memory(GiB)": 33.01,
+      "step": 2465,
+      "train_speed(iter/s)": 0.192942
+    },
+    {
+      "acc": 0.86496525,
+      "epoch": 0.8184227965540093,
+      "grad_norm": 7.462728624822657,
+      "learning_rate": 9.749915794085862e-06,
+      "loss": 0.62141933,
+      "memory(GiB)": 33.01,
+      "step": 2470,
+      "train_speed(iter/s)": 0.192946
+    },
+    {
+      "acc": 0.8760519,
+      "epoch": 0.820079522862823,
+      "grad_norm": 5.589536885979733,
+      "learning_rate": 9.752439704638696e-06,
+      "loss": 0.6348043,
+      "memory(GiB)": 33.01,
+      "step": 2475,
+      "train_speed(iter/s)": 0.192949
+    },
+    {
+      "acc": 0.86904612,
+      "epoch": 0.8217362491716369,
+      "grad_norm": 3.553935619178476,
+      "learning_rate": 9.754958521523983e-06,
+      "loss": 0.65660267,
+      "memory(GiB)": 33.01,
+      "step": 2480,
+      "train_speed(iter/s)": 0.192954
+    },
+    {
+      "acc": 0.86920958,
+      "epoch": 0.8233929754804507,
+      "grad_norm": 8.160696258419575,
+      "learning_rate": 9.757472265260047e-06,
+      "loss": 0.63985438,
+      "memory(GiB)": 33.01,
+      "step": 2485,
+      "train_speed(iter/s)": 0.192957
+    },
+    {
+      "acc": 0.86231823,
+      "epoch": 0.8250497017892644,
+      "grad_norm": 7.278889798912907,
+      "learning_rate": 9.759980956241476e-06,
+      "loss": 0.6573843,
+      "memory(GiB)": 33.01,
+      "step": 2490,
+      "train_speed(iter/s)": 0.192962
+    },
+    {
+      "acc": 0.88577518,
+      "epoch": 0.8267064280980781,
+      "grad_norm": 2.5005106896064127,
+      "learning_rate": 9.762484614740126e-06,
+      "loss": 0.57792425,
+      "memory(GiB)": 33.01,
+      "step": 2495,
+      "train_speed(iter/s)": 0.192965
+    },
+    {
+      "acc": 0.88653049,
+      "epoch": 0.828363154406892,
+      "grad_norm": 3.5019587588066097,
+      "learning_rate": 9.764983260906098e-06,
+      "loss": 0.61271315,
+      "memory(GiB)": 33.01,
+      "step": 2500,
+      "train_speed(iter/s)": 0.192969
+    },
+    {
+      "acc": 0.8679534,
+      "epoch": 0.8300198807157058,
+      "grad_norm": 3.728792830280097,
+      "learning_rate": 9.767476914768714e-06,
+      "loss": 0.64312353,
+      "memory(GiB)": 33.01,
+      "step": 2505,
+      "train_speed(iter/s)": 0.192973
+    },
+    {
+      "acc": 0.87462015,
+      "epoch": 0.8316766070245195,
+      "grad_norm": 4.3700933875496135,
+      "learning_rate": 9.769965596237475e-06,
+      "loss": 0.60275097,
+      "memory(GiB)": 33.01,
+      "step": 2510,
+      "train_speed(iter/s)": 0.192976
+    },
+    {
+      "acc": 0.87128067,
+      "epoch": 0.8333333333333334,
+      "grad_norm": 7.612978714975665,
+      "learning_rate": 9.772449325103028e-06,
+      "loss": 0.65022855,
+      "memory(GiB)": 33.01,
+      "step": 2515,
+      "train_speed(iter/s)": 0.19298
+    },
+    {
+      "acc": 0.8678833,
+      "epoch": 0.8349900596421471,
+      "grad_norm": 7.616382306500804,
+      "learning_rate": 9.774928121038095e-06,
+      "loss": 0.64670591,
+      "memory(GiB)": 33.01,
+      "step": 2520,
+      "train_speed(iter/s)": 0.192983
+    },
+    {
+      "acc": 0.87310963,
+      "epoch": 0.8366467859509609,
+      "grad_norm": 3.3425387410443386,
+      "learning_rate": 9.777402003598412e-06,
+      "loss": 0.63883715,
+      "memory(GiB)": 33.01,
+      "step": 2525,
+      "train_speed(iter/s)": 0.192987
+    },
+    {
+      "acc": 0.8770895,
+      "epoch": 0.8383035122597747,
+      "grad_norm": 8.563408061589213,
+      "learning_rate": 9.779870992223668e-06,
+      "loss": 0.62872858,
+      "memory(GiB)": 33.01,
+      "step": 2530,
+      "train_speed(iter/s)": 0.192992
+    },
+    {
+      "acc": 0.87358456,
+      "epoch": 0.8399602385685885,
+      "grad_norm": 2.963533578178825,
+      "learning_rate": 9.782335106238396e-06,
+      "loss": 0.64085312,
+      "memory(GiB)": 33.01,
+      "step": 2535,
+      "train_speed(iter/s)": 0.192996
+    },
+    {
+      "acc": 0.87006836,
+      "epoch": 0.8416169648774022,
+      "grad_norm": 6.722505279492165,
+      "learning_rate": 9.784794364852908e-06,
+      "loss": 0.63312893,
+      "memory(GiB)": 33.01,
+      "step": 2540,
+      "train_speed(iter/s)": 0.193
+    },
+    {
+      "acc": 0.87112741,
+      "epoch": 0.8432736911862161,
+      "grad_norm": 3.870915323241978,
+      "learning_rate": 9.78724878716417e-06,
+      "loss": 0.61600509,
+      "memory(GiB)": 33.01,
+      "step": 2545,
+      "train_speed(iter/s)": 0.193003
+    },
+    {
+      "acc": 0.86981926,
+      "epoch": 0.8449304174950298,
+      "grad_norm": 7.411035478079732,
+      "learning_rate": 9.789698392156712e-06,
+      "loss": 0.62298193,
+      "memory(GiB)": 33.01,
+      "step": 2550,
+      "train_speed(iter/s)": 0.193007
+    },
+    {
+      "acc": 0.88824244,
+      "epoch": 0.8465871438038436,
+      "grad_norm": 5.175288938119396,
+      "learning_rate": 9.792143198703494e-06,
+      "loss": 0.57580199,
+      "memory(GiB)": 33.01,
+      "step": 2555,
+      "train_speed(iter/s)": 0.19301
+    },
+    {
+      "acc": 0.87506332,
+      "epoch": 0.8482438701126573,
+      "grad_norm": 3.0398547056625933,
+      "learning_rate": 9.794583225566784e-06,
+      "loss": 0.61144085,
+      "memory(GiB)": 33.01,
+      "step": 2560,
+      "train_speed(iter/s)": 0.193014
+    },
+    {
+      "acc": 0.88642302,
+      "epoch": 0.8499005964214712,
+      "grad_norm": 3.400815807481571,
+      "learning_rate": 9.797018491399024e-06,
+      "loss": 0.59812255,
+      "memory(GiB)": 33.01,
+      "step": 2565,
+      "train_speed(iter/s)": 0.193018
+    },
+    {
+      "acc": 0.89076796,
+      "epoch": 0.851557322730285,
+      "grad_norm": 5.2301657506873855,
+      "learning_rate": 9.799449014743683e-06,
+      "loss": 0.56942215,
+      "memory(GiB)": 33.01,
+      "step": 2570,
+      "train_speed(iter/s)": 0.193022
+    },
+    {
+      "acc": 0.87888985,
+      "epoch": 0.8532140490390987,
+      "grad_norm": 3.7446013107384593,
+      "learning_rate": 9.801874814036104e-06,
+      "loss": 0.59126511,
+      "memory(GiB)": 33.01,
+      "step": 2575,
+      "train_speed(iter/s)": 0.193026
+    },
+    {
+      "acc": 0.87289181,
+      "epoch": 0.8548707753479126,
+      "grad_norm": 7.364936938315251,
+      "learning_rate": 9.804295907604344e-06,
+      "loss": 0.68985062,
+      "memory(GiB)": 33.01,
+      "step": 2580,
+      "train_speed(iter/s)": 0.193029
+    },
+    {
+      "acc": 0.88986206,
+      "epoch": 0.8565275016567263,
+      "grad_norm": 5.180863897320667,
+      "learning_rate": 9.806712313670004e-06,
+      "loss": 0.54514599,
+      "memory(GiB)": 33.01,
+      "step": 2585,
+      "train_speed(iter/s)": 0.193033
+    },
+    {
+      "acc": 0.88568058,
+      "epoch": 0.8581842279655401,
+      "grad_norm": 3.9336394530292336,
+      "learning_rate": 9.809124050349054e-06,
+      "loss": 0.57606583,
+      "memory(GiB)": 33.01,
+      "step": 2590,
+      "train_speed(iter/s)": 0.193037
+    },
+    {
+      "acc": 0.8778533,
+      "epoch": 0.8598409542743539,
+      "grad_norm": 5.933507306693597,
+      "learning_rate": 9.811531135652644e-06,
+      "loss": 0.6583971,
+      "memory(GiB)": 33.01,
+      "step": 2595,
+      "train_speed(iter/s)": 0.193041
+    },
+    {
+      "acc": 0.87049732,
+      "epoch": 0.8614976805831677,
+      "grad_norm": 8.048986545194238,
+      "learning_rate": 9.813933587487917e-06,
+      "loss": 0.63537345,
+      "memory(GiB)": 33.01,
+      "step": 2600,
+      "train_speed(iter/s)": 0.193045
+    },
+    {
+      "acc": 0.88129292,
+      "epoch": 0.8631544068919814,
+      "grad_norm": 8.071277659969068,
+      "learning_rate": 9.8163314236588e-06,
+      "loss": 0.60130939,
+      "memory(GiB)": 33.01,
+      "step": 2605,
+      "train_speed(iter/s)": 0.193049
+    },
+    {
+      "acc": 0.86977844,
+      "epoch": 0.8648111332007953,
+      "grad_norm": 3.8453891682437282,
+      "learning_rate": 9.818724661866802e-06,
+      "loss": 0.60652094,
+      "memory(GiB)": 33.01,
+      "step": 2610,
+      "train_speed(iter/s)": 0.193052
+    },
+    {
+      "acc": 0.8910141,
+      "epoch": 0.866467859509609,
+      "grad_norm": 3.9126981491650925,
+      "learning_rate": 9.821113319711799e-06,
+      "loss": 0.57247858,
+      "memory(GiB)": 33.01,
+      "step": 2615,
+      "train_speed(iter/s)": 0.193056
+    },
+    {
+      "acc": 0.88858194,
+      "epoch": 0.8681245858184228,
+      "grad_norm": 4.436881308598289,
+      "learning_rate": 9.823497414692806e-06,
+      "loss": 0.54993615,
+      "memory(GiB)": 33.01,
+      "step": 2620,
+      "train_speed(iter/s)": 0.19306
+    },
+    {
+      "acc": 0.89008675,
+      "epoch": 0.8697813121272365,
+      "grad_norm": 7.166566696540512,
+      "learning_rate": 9.825876964208749e-06,
+      "loss": 0.61355448,
+      "memory(GiB)": 33.01,
+      "step": 2625,
+      "train_speed(iter/s)": 0.193063
+    },
+    {
+      "acc": 0.87891541,
+      "epoch": 0.8714380384360504,
+      "grad_norm": 4.395432702419088,
+      "learning_rate": 9.828251985559222e-06,
+      "loss": 0.61320391,
+      "memory(GiB)": 33.01,
+      "step": 2630,
+      "train_speed(iter/s)": 0.193067
+    },
+    {
+      "acc": 0.87204447,
+      "epoch": 0.8730947647448641,
+      "grad_norm": 3.3348825245813267,
+      "learning_rate": 9.830622495945252e-06,
+      "loss": 0.60965052,
+      "memory(GiB)": 33.01,
+      "step": 2635,
+      "train_speed(iter/s)": 0.19307
+    },
+    {
+      "acc": 0.88088131,
+      "epoch": 0.8747514910536779,
+      "grad_norm": 3.9816964216160633,
+      "learning_rate": 9.832988512470033e-06,
+      "loss": 0.57345366,
+      "memory(GiB)": 33.01,
+      "step": 2640,
+      "train_speed(iter/s)": 0.193074
+    },
+    {
+      "acc": 0.88114119,
+      "epoch": 0.8764082173624917,
+      "grad_norm": 9.249511642840648,
+      "learning_rate": 9.835350052139685e-06,
+      "loss": 0.58579946,
+      "memory(GiB)": 33.01,
+      "step": 2645,
+      "train_speed(iter/s)": 0.193078
+    },
+    {
+      "acc": 0.87508268,
+      "epoch": 0.8780649436713055,
+      "grad_norm": 3.0565908923917724,
+      "learning_rate": 9.837707131863961e-06,
+      "loss": 0.65310869,
+      "memory(GiB)": 33.01,
+      "step": 2650,
+      "train_speed(iter/s)": 0.193082
+    },
+    {
+      "acc": 0.86386747,
+      "epoch": 0.8797216699801192,
+      "grad_norm": 4.686453360827556,
+      "learning_rate": 9.840059768457001e-06,
+      "loss": 0.63752556,
+      "memory(GiB)": 33.01,
+      "step": 2655,
+      "train_speed(iter/s)": 0.193085
+    },
+    {
+      "acc": 0.86603851,
+      "epoch": 0.8813783962889331,
+      "grad_norm": 5.402025133939387,
+      "learning_rate": 9.842407978638033e-06,
+      "loss": 0.67756987,
+      "memory(GiB)": 33.01,
+      "step": 2660,
+      "train_speed(iter/s)": 0.193075
+    },
+    {
+      "acc": 0.88185043,
+      "epoch": 0.8830351225977469,
+      "grad_norm": 3.2103146092290977,
+      "learning_rate": 9.844751779032092e-06,
+      "loss": 0.61673317,
+      "memory(GiB)": 33.01,
+      "step": 2665,
+      "train_speed(iter/s)": 0.193075
+    },
+    {
+      "acc": 0.86560764,
+      "epoch": 0.8846918489065606,
+      "grad_norm": 8.33317075660369,
+      "learning_rate": 9.847091186170731e-06,
+      "loss": 0.68903618,
+      "memory(GiB)": 33.01,
+      "step": 2670,
+      "train_speed(iter/s)": 0.193079
+    },
+    {
+      "acc": 0.87715178,
+      "epoch": 0.8863485752153745,
+      "grad_norm": 4.223430736170169,
+      "learning_rate": 9.849426216492708e-06,
+      "loss": 0.63275361,
+      "memory(GiB)": 33.01,
+      "step": 2675,
+      "train_speed(iter/s)": 0.193083
+    },
+    {
+      "acc": 0.87970896,
+      "epoch": 0.8880053015241882,
+      "grad_norm": 2.94121098627528,
+      "learning_rate": 9.851756886344694e-06,
+      "loss": 0.57373309,
+      "memory(GiB)": 33.01,
+      "step": 2680,
+      "train_speed(iter/s)": 0.193086
+    },
+    {
+      "acc": 0.88578835,
+      "epoch": 0.889662027833002,
+      "grad_norm": 3.300063503820921,
+      "learning_rate": 9.854083211981951e-06,
+      "loss": 0.58510609,
+      "memory(GiB)": 33.01,
+      "step": 2685,
+      "train_speed(iter/s)": 0.193089
+    },
+    {
+      "acc": 0.87865191,
+      "epoch": 0.8913187541418157,
+      "grad_norm": 3.9845165580069737,
+      "learning_rate": 9.85640520956901e-06,
+      "loss": 0.57173872,
+      "memory(GiB)": 33.01,
+      "step": 2690,
+      "train_speed(iter/s)": 0.193092
+    },
+    {
+      "acc": 0.86326895,
+      "epoch": 0.8929754804506296,
+      "grad_norm": 6.948348088961666,
+      "learning_rate": 9.858722895180354e-06,
+      "loss": 0.68775687,
+      "memory(GiB)": 33.01,
+      "step": 2695,
+      "train_speed(iter/s)": 0.193096
+    },
+    {
+      "acc": 0.87145214,
+      "epoch": 0.8946322067594433,
+      "grad_norm": 8.752287203643892,
+      "learning_rate": 9.861036284801077e-06,
+      "loss": 0.61272125,
+      "memory(GiB)": 33.01,
+      "step": 2700,
+      "train_speed(iter/s)": 0.193099
+    },
+    {
+      "acc": 0.87385406,
+      "epoch": 0.8962889330682571,
+      "grad_norm": 6.90240422582082,
+      "learning_rate": 9.863345394327551e-06,
+      "loss": 0.66489458,
+      "memory(GiB)": 33.01,
+      "step": 2705,
+      "train_speed(iter/s)": 0.193103
+    },
+    {
+      "acc": 0.88107843,
+      "epoch": 0.8979456593770709,
+      "grad_norm": 2.616336324483754,
+      "learning_rate": 9.86565023956808e-06,
+      "loss": 0.59428682,
+      "memory(GiB)": 33.01,
+      "step": 2710,
+      "train_speed(iter/s)": 0.193106
+    },
+    {
+      "acc": 0.88717527,
+      "epoch": 0.8996023856858847,
+      "grad_norm": 5.794634490985694,
+      "learning_rate": 9.867950836243542e-06,
+      "loss": 0.57184386,
+      "memory(GiB)": 33.01,
+      "step": 2715,
+      "train_speed(iter/s)": 0.193109
+    },
+    {
+      "acc": 0.87975388,
+      "epoch": 0.9012591119946984,
+      "grad_norm": 5.012366580025711,
+      "learning_rate": 9.870247199988051e-06,
+      "loss": 0.58441563,
+      "memory(GiB)": 33.01,
+      "step": 2720,
+      "train_speed(iter/s)": 0.193113
+    },
+    {
+      "acc": 0.88116417,
+      "epoch": 0.9029158383035123,
+      "grad_norm": 2.9602394894892443,
+      "learning_rate": 9.87253934634957e-06,
+      "loss": 0.58528523,
+      "memory(GiB)": 33.01,
+      "step": 2725,
+      "train_speed(iter/s)": 0.193116
+    },
+    {
+      "acc": 0.874856,
+      "epoch": 0.904572564612326,
+      "grad_norm": 5.593801991819785,
+      "learning_rate": 9.874827290790568e-06,
+      "loss": 0.63316264,
+      "memory(GiB)": 33.01,
+      "step": 2730,
+      "train_speed(iter/s)": 0.193119
+    },
+    {
+      "acc": 0.87478542,
+      "epoch": 0.9062292909211398,
+      "grad_norm": 2.853767232052871,
+      "learning_rate": 9.877111048688618e-06,
+      "loss": 0.63812952,
+      "memory(GiB)": 33.01,
+      "step": 2735,
+      "train_speed(iter/s)": 0.193122
+    },
+    {
+      "acc": 0.87874794,
+      "epoch": 0.9078860172299537,
+      "grad_norm": 4.479632406548384,
+      "learning_rate": 9.879390635337044e-06,
+      "loss": 0.64914513,
+      "memory(GiB)": 33.01,
+      "step": 2740,
+      "train_speed(iter/s)": 0.193125
+    },
+    {
+      "acc": 0.87867851,
+      "epoch": 0.9095427435387674,
+      "grad_norm": 3.9428390634160415,
+      "learning_rate": 9.881666065945522e-06,
+      "loss": 0.60804138,
+      "memory(GiB)": 33.01,
+      "step": 2745,
+      "train_speed(iter/s)": 0.193129
+    },
+    {
+      "acc": 0.8855443,
+      "epoch": 0.9111994698475812,
+      "grad_norm": 4.549793534128653,
+      "learning_rate": 9.883937355640688e-06,
+      "loss": 0.59703541,
+      "memory(GiB)": 33.01,
+      "step": 2750,
+      "train_speed(iter/s)": 0.193132
+    },
+    {
+      "acc": 0.88780842,
+      "epoch": 0.912856196156395,
+      "grad_norm": 3.1111898746033857,
+      "learning_rate": 9.88620451946674e-06,
+      "loss": 0.59631329,
+      "memory(GiB)": 33.01,
+      "step": 2755,
+      "train_speed(iter/s)": 0.193135
+    },
+    {
+      "acc": 0.88322296,
+      "epoch": 0.9145129224652088,
+      "grad_norm": 4.568535516335659,
+      "learning_rate": 9.88846757238605e-06,
+      "loss": 0.6033711,
+      "memory(GiB)": 33.01,
+      "step": 2760,
+      "train_speed(iter/s)": 0.193139
+    },
+    {
+      "acc": 0.88480091,
+      "epoch": 0.9161696487740225,
+      "grad_norm": 3.13589447383103,
+      "learning_rate": 9.890726529279738e-06,
+      "loss": 0.58900795,
+      "memory(GiB)": 33.01,
+      "step": 2765,
+      "train_speed(iter/s)": 0.193141
+    },
+    {
+      "acc": 0.87481556,
+      "epoch": 0.9178263750828363,
+      "grad_norm": 5.493848696205705,
+      "learning_rate": 9.89298140494827e-06,
+      "loss": 0.6229105,
+      "memory(GiB)": 33.01,
+      "step": 2770,
+      "train_speed(iter/s)": 0.193145
+    },
+    {
+      "acc": 0.88898182,
+      "epoch": 0.9194831013916501,
+      "grad_norm": 5.590150231287597,
+      "learning_rate": 9.895232214112037e-06,
+      "loss": 0.56973162,
+      "memory(GiB)": 33.01,
+      "step": 2775,
+      "train_speed(iter/s)": 0.193148
+    },
+    {
+      "acc": 0.87322502,
+      "epoch": 0.9211398277004639,
+      "grad_norm": 3.4763799314446353,
+      "learning_rate": 9.897478971411934e-06,
+      "loss": 0.61146326,
+      "memory(GiB)": 33.01,
+      "step": 2780,
+      "train_speed(iter/s)": 0.193151
+    },
+    {
+      "acc": 0.88107433,
+      "epoch": 0.9227965540092776,
+      "grad_norm": 3.8196884536175557,
+      "learning_rate": 9.899721691409925e-06,
+      "loss": 0.58410368,
+      "memory(GiB)": 33.01,
+      "step": 2785,
+      "train_speed(iter/s)": 0.193155
+    },
+    {
+      "acc": 0.88541393,
+      "epoch": 0.9244532803180915,
+      "grad_norm": 6.503850716397312,
+      "learning_rate": 9.901960388589616e-06,
+      "loss": 0.56620717,
+      "memory(GiB)": 33.01,
+      "step": 2790,
+      "train_speed(iter/s)": 0.193158
+    },
+    {
+      "acc": 0.89066582,
+      "epoch": 0.9261100066269052,
+      "grad_norm": 4.407683320212889,
+      "learning_rate": 9.904195077356816e-06,
+      "loss": 0.55348973,
+      "memory(GiB)": 33.01,
+      "step": 2795,
+      "train_speed(iter/s)": 0.19316
+    },
+    {
+      "acc": 0.87991524,
+      "epoch": 0.927766732935719,
+      "grad_norm": 12.359804406687395,
+      "learning_rate": 9.906425772040086e-06,
+      "loss": 0.57670903,
+      "memory(GiB)": 33.01,
+      "step": 2800,
+      "train_speed(iter/s)": 0.193163
+    },
+    {
+      "acc": 0.87292213,
+      "epoch": 0.9294234592445328,
+      "grad_norm": 3.5608286359015193,
+      "learning_rate": 9.9086524868913e-06,
+      "loss": 0.62728844,
+      "memory(GiB)": 33.01,
+      "step": 2805,
+      "train_speed(iter/s)": 0.193166
+    },
+    {
+      "acc": 0.8768137,
+      "epoch": 0.9310801855533466,
+      "grad_norm": 4.492959214169583,
+      "learning_rate": 9.910875236086185e-06,
+      "loss": 0.59196949,
+      "memory(GiB)": 33.01,
+      "step": 2810,
+      "train_speed(iter/s)": 0.19317
+    },
+    {
+      "acc": 0.87537785,
+      "epoch": 0.9327369118621603,
+      "grad_norm": 6.076829852184719,
+      "learning_rate": 9.913094033724863e-06,
+      "loss": 0.61625524,
+      "memory(GiB)": 33.01,
+      "step": 2815,
+      "train_speed(iter/s)": 0.193173
+    },
+    {
+      "acc": 0.87883272,
+      "epoch": 0.9343936381709742,
+      "grad_norm": 5.17410281157418,
+      "learning_rate": 9.915308893832387e-06,
+      "loss": 0.61370792,
+      "memory(GiB)": 33.01,
+      "step": 2820,
+      "train_speed(iter/s)": 0.193176
+    },
+    {
+      "acc": 0.88947096,
+      "epoch": 0.936050364479788,
+      "grad_norm": 2.354419652715678,
+      "learning_rate": 9.917519830359278e-06,
+      "loss": 0.56221933,
+      "memory(GiB)": 33.01,
+      "step": 2825,
+      "train_speed(iter/s)": 0.19318
+    },
+    {
+      "acc": 0.8921813,
+      "epoch": 0.9377070907886017,
+      "grad_norm": 2.3352101015851763,
+      "learning_rate": 9.91972685718205e-06,
+      "loss": 0.5675477,
+      "memory(GiB)": 33.01,
+      "step": 2830,
+      "train_speed(iter/s)": 0.193183
+    },
+    {
+      "acc": 0.89346809,
+      "epoch": 0.9393638170974155,
+      "grad_norm": 3.8779157237954567,
+      "learning_rate": 9.921929988103727e-06,
+      "loss": 0.55961146,
+      "memory(GiB)": 33.01,
+      "step": 2835,
+      "train_speed(iter/s)": 0.193186
+    },
+    {
+      "acc": 0.88623056,
+      "epoch": 0.9410205434062293,
+      "grad_norm": 7.307612857800359,
+      "learning_rate": 9.92412923685437e-06,
+      "loss": 0.59642801,
+      "memory(GiB)": 33.01,
+      "step": 2840,
+      "train_speed(iter/s)": 0.193189
+    },
+    {
+      "acc": 0.88649006,
+      "epoch": 0.9426772697150431,
+      "grad_norm": 3.5190311837873796,
+      "learning_rate": 9.926324617091579e-06,
+      "loss": 0.5734262,
+      "memory(GiB)": 33.01,
+      "step": 2845,
+      "train_speed(iter/s)": 0.193191
+    },
+    {
+      "acc": 0.8924675,
+      "epoch": 0.9443339960238568,
+      "grad_norm": 8.106329865969013,
+      "learning_rate": 9.928516142401016e-06,
+      "loss": 0.56243801,
+      "memory(GiB)": 33.01,
+      "step": 2850,
+      "train_speed(iter/s)": 0.193195
+    },
+    {
+      "acc": 0.88808842,
+      "epoch": 0.9459907223326707,
+      "grad_norm": 3.559472198524817,
+      "learning_rate": 9.930703826296901e-06,
+      "loss": 0.53762207,
+      "memory(GiB)": 33.01,
+      "step": 2855,
+      "train_speed(iter/s)": 0.193198
+    },
+    {
+      "acc": 0.88387766,
+      "epoch": 0.9476474486414844,
+      "grad_norm": 3.4791848519540687,
+      "learning_rate": 9.932887682222506e-06,
+      "loss": 0.54636021,
+      "memory(GiB)": 33.01,
+      "step": 2860,
+      "train_speed(iter/s)": 0.193202
+    },
+    {
+      "acc": 0.88861408,
+      "epoch": 0.9493041749502982,
+      "grad_norm": 4.650097463223281,
+      "learning_rate": 9.935067723550667e-06,
+      "loss": 0.61271467,
+      "memory(GiB)": 33.01,
+      "step": 2865,
+      "train_speed(iter/s)": 0.193205
+    },
+    {
+      "acc": 0.89421978,
+      "epoch": 0.950960901259112,
+      "grad_norm": 3.9924528892139226,
+      "learning_rate": 9.937243963584262e-06,
+      "loss": 0.54757972,
+      "memory(GiB)": 33.01,
+      "step": 2870,
+      "train_speed(iter/s)": 0.193208
+    },
+    {
+      "acc": 0.89381542,
+      "epoch": 0.9526176275679258,
+      "grad_norm": 4.680966813328855,
+      "learning_rate": 9.939416415556704e-06,
+      "loss": 0.5293014,
+      "memory(GiB)": 33.01,
+      "step": 2875,
+      "train_speed(iter/s)": 0.193211
+    },
+    {
+      "acc": 0.89750099,
+      "epoch": 0.9542743538767395,
+      "grad_norm": 6.0306331010496805,
+      "learning_rate": 9.941585092632416e-06,
+      "loss": 0.53185587,
+      "memory(GiB)": 33.01,
+      "step": 2880,
+      "train_speed(iter/s)": 0.193213
+    },
+    {
+      "acc": 0.88830643,
+      "epoch": 0.9559310801855534,
+      "grad_norm": 9.350386874260305,
+      "learning_rate": 9.943750007907325e-06,
+      "loss": 0.58860159,
+      "memory(GiB)": 33.01,
+      "step": 2885,
+      "train_speed(iter/s)": 0.193205
+    },
+    {
+      "acc": 0.87832508,
+      "epoch": 0.9575878064943671,
+      "grad_norm": 2.981169782948862,
+      "learning_rate": 9.945911174409318e-06,
+      "loss": 0.59479094,
+      "memory(GiB)": 33.01,
+      "step": 2890,
+      "train_speed(iter/s)": 0.193203
+    },
+    {
+      "acc": 0.89966698,
+      "epoch": 0.9592445328031809,
+      "grad_norm": 4.547161210484024,
+      "learning_rate": 9.94806860509873e-06,
+      "loss": 0.53411198,
+      "memory(GiB)": 33.01,
+      "step": 2895,
+      "train_speed(iter/s)": 0.193203
+    },
+    {
+      "acc": 0.89919624,
+      "epoch": 0.9609012591119946,
+      "grad_norm": 5.715099384971192,
+      "learning_rate": 9.950222312868795e-06,
+      "loss": 0.54964809,
+      "memory(GiB)": 33.01,
+      "step": 2900,
+      "train_speed(iter/s)": 0.193205
+    },
+    {
+      "acc": 0.89024811,
+      "epoch": 0.9625579854208085,
+      "grad_norm": 15.890328578611772,
+      "learning_rate": 9.952372310546117e-06,
+      "loss": 0.53085089,
+      "memory(GiB)": 33.01,
+      "step": 2905,
+      "train_speed(iter/s)": 0.193209
+    },
+    {
+      "acc": 0.89768763,
+      "epoch": 0.9642147117296223,
+      "grad_norm": 3.423345092171575,
+      "learning_rate": 9.954518610891133e-06,
+      "loss": 0.51252246,
+      "memory(GiB)": 33.01,
+      "step": 2910,
+      "train_speed(iter/s)": 0.193212
+    },
+    {
+      "acc": 0.89978552,
+      "epoch": 0.965871438038436,
+      "grad_norm": 3.4753424048647212,
+      "learning_rate": 9.956661226598551e-06,
+      "loss": 0.50663137,
+      "memory(GiB)": 33.01,
+      "step": 2915,
+      "train_speed(iter/s)": 0.193215
+    },
+    {
+      "acc": 0.88726511,
+      "epoch": 0.9675281643472499,
+      "grad_norm": 8.177323401447397,
+      "learning_rate": 9.958800170297816e-06,
+      "loss": 0.57537966,
+      "memory(GiB)": 33.01,
+      "step": 2920,
+      "train_speed(iter/s)": 0.193218
+    },
+    {
+      "acc": 0.88609066,
+      "epoch": 0.9691848906560636,
+      "grad_norm": 4.191968466292702,
+      "learning_rate": 9.96093545455355e-06,
+      "loss": 0.58137269,
+      "memory(GiB)": 33.01,
+      "step": 2925,
+      "train_speed(iter/s)": 0.193221
+    },
+    {
+      "acc": 0.88744173,
+      "epoch": 0.9708416169648774,
+      "grad_norm": 5.809899061291322,
+      "learning_rate": 9.963067091865996e-06,
+      "loss": 0.56933103,
+      "memory(GiB)": 33.01,
+      "step": 2930,
+      "train_speed(iter/s)": 0.193224
+    },
+    {
+      "acc": 0.88377037,
+      "epoch": 0.9724983432736912,
+      "grad_norm": 3.0688050587103826,
+      "learning_rate": 9.965195094671458e-06,
+      "loss": 0.57009926,
+      "memory(GiB)": 33.01,
+      "step": 2935,
+      "train_speed(iter/s)": 0.193227
+    },
+    {
+      "acc": 0.88039722,
+      "epoch": 0.974155069582505,
+      "grad_norm": 2.7636680800360995,
+      "learning_rate": 9.967319475342736e-06,
+      "loss": 0.54072194,
+      "memory(GiB)": 33.01,
+      "step": 2940,
+      "train_speed(iter/s)": 0.19323
+    },
+    {
+      "acc": 0.88092518,
+      "epoch": 0.9758117958913187,
+      "grad_norm": 2.642166247104512,
+      "learning_rate": 9.969440246189554e-06,
+      "loss": 0.56847663,
+      "memory(GiB)": 33.01,
+      "step": 2945,
+      "train_speed(iter/s)": 0.193233
+    },
+    {
+      "acc": 0.88988876,
+      "epoch": 0.9774685222001326,
+      "grad_norm": 3.2327427727920437,
+      "learning_rate": 9.971557419458993e-06,
+      "loss": 0.53624258,
+      "memory(GiB)": 33.01,
+      "step": 2950,
+      "train_speed(iter/s)": 0.193236
+    },
+    {
+      "acc": 0.8884038,
+      "epoch": 0.9791252485089463,
+      "grad_norm": 5.918917242589188,
+      "learning_rate": 9.973671007335915e-06,
+      "loss": 0.53204603,
+      "memory(GiB)": 33.01,
+      "step": 2955,
+      "train_speed(iter/s)": 0.19324
+    },
+    {
+      "acc": 0.88612843,
+      "epoch": 0.9807819748177601,
+      "grad_norm": 3.5210439396485036,
+      "learning_rate": 9.975781021943376e-06,
+      "loss": 0.5242445,
+      "memory(GiB)": 33.01,
+      "step": 2960,
+      "train_speed(iter/s)": 0.193243
+    },
+    {
+      "acc": 0.88615685,
+      "epoch": 0.9824387011265739,
+      "grad_norm": 3.62728999755055,
+      "learning_rate": 9.977887475343056e-06,
+      "loss": 0.51521478,
+      "memory(GiB)": 33.01,
+      "step": 2965,
+      "train_speed(iter/s)": 0.193246
+    },
+    {
+      "acc": 0.89091187,
+      "epoch": 0.9840954274353877,
+      "grad_norm": 4.920774391651594,
+      "learning_rate": 9.979990379535667e-06,
+      "loss": 0.56898975,
+      "memory(GiB)": 33.01,
+      "step": 2970,
+      "train_speed(iter/s)": 0.193249
+    },
+    {
+      "acc": 0.89261436,
+      "epoch": 0.9857521537442014,
+      "grad_norm": 3.3201174498123835,
+      "learning_rate": 9.982089746461353e-06,
+      "loss": 0.53635111,
+      "memory(GiB)": 33.01,
+      "step": 2975,
+      "train_speed(iter/s)": 0.193252
+    },
+    {
+      "acc": 0.88731613,
+      "epoch": 0.9874088800530152,
+      "grad_norm": 5.747764318444423,
+      "learning_rate": 9.984185588000118e-06,
+      "loss": 0.60284166,
+      "memory(GiB)": 33.01,
+      "step": 2980,
+      "train_speed(iter/s)": 0.193255
+    },
+    {
+      "acc": 0.88962011,
+      "epoch": 0.989065606361829,
+      "grad_norm": 3.864722841804485,
+      "learning_rate": 9.986277915972207e-06,
+      "loss": 0.54507532,
+      "memory(GiB)": 33.01,
+      "step": 2985,
+      "train_speed(iter/s)": 0.193258
+    },
+    {
+      "acc": 0.89195766,
+      "epoch": 0.9907223326706428,
+      "grad_norm": 6.5018775023143265,
+      "learning_rate": 9.988366742138523e-06,
+      "loss": 0.52294307,
+      "memory(GiB)": 33.01,
+      "step": 2990,
+      "train_speed(iter/s)": 0.19326
+    },
+    {
+      "acc": 0.89480324,
+      "epoch": 0.9923790589794566,
+      "grad_norm": 7.012731743450955,
+      "learning_rate": 9.990452078201013e-06,
+      "loss": 0.49354048,
+      "memory(GiB)": 33.01,
+      "step": 2995,
+      "train_speed(iter/s)": 0.193263
+    },
+    {
+      "acc": 0.89031048,
+      "epoch": 0.9940357852882704,
+      "grad_norm": 8.816750683995743,
+      "learning_rate": 9.992533935803069e-06,
+      "loss": 0.54319358,
+      "memory(GiB)": 33.01,
+      "step": 3000,
+      "train_speed(iter/s)": 0.193265
+    },
+    {
+      "acc": 0.89605188,
+      "epoch": 0.9956925115970842,
+      "grad_norm": 3.7296249915122592,
+      "learning_rate": 9.994612326529912e-06,
+      "loss": 0.5190588,
+      "memory(GiB)": 33.01,
+      "step": 3005,
+      "train_speed(iter/s)": 0.193268
+    },
+    {
+      "acc": 0.90042381,
+      "epoch": 0.9973492379058979,
+      "grad_norm": 4.378739933514348,
+      "learning_rate": 9.996687261908986e-06,
+      "loss": 0.48994136,
+      "memory(GiB)": 33.01,
+      "step": 3010,
+      "train_speed(iter/s)": 0.193271
+    },
+    {
+      "acc": 0.90874329,
+      "epoch": 0.9990059642147118,
+      "grad_norm": 3.503144553319389,
+      "learning_rate": 9.998758753410328e-06,
+      "loss": 0.45982637,
+      "memory(GiB)": 33.01,
+      "step": 3015,
+      "train_speed(iter/s)": 0.193274
+    },
+    {
+      "epoch": 1.0,
+      "eval_acc": 0.9013307583879955,
+      "eval_loss": 0.24342016875743866,
+      "eval_runtime": 504.8587,
+      "eval_samples_per_second": 10.627,
+      "eval_steps_per_second": 1.329,
+      "step": 3018
+    },
+    {
+      "acc": 0.90102835,
+      "epoch": 1.0006626905235254,
+      "grad_norm": 2.532139645350858,
+      "learning_rate": 9.999999969986915e-06,
+      "loss": 0.49257302,
+      "memory(GiB)": 33.01,
+      "step": 3020,
+      "train_speed(iter/s)": 0.186613
+    },
+    {
+      "acc": 0.90085068,
+      "epoch": 1.0023194168323393,
+      "grad_norm": 4.620805847378081,
+      "learning_rate": 9.99999963233971e-06,
+      "loss": 0.47919602,
+      "memory(GiB)": 33.01,
+      "step": 3025,
+      "train_speed(iter/s)": 0.186623
+    },
+    {
+      "acc": 0.89500475,
+      "epoch": 1.0039761431411531,
+      "grad_norm": 7.341574928547011,
+      "learning_rate": 9.999998919528973e-06,
+      "loss": 0.51861906,
+      "memory(GiB)": 33.01,
+      "step": 3030,
+      "train_speed(iter/s)": 0.186631
+    },
+    {
+      "acc": 0.89477425,
+      "epoch": 1.0056328694499668,
+      "grad_norm": 5.892171722977981,
+      "learning_rate": 9.999997831554753e-06,
+      "loss": 0.55375786,
+      "memory(GiB)": 33.01,
+      "step": 3035,
+      "train_speed(iter/s)": 0.18664
+    },
+    {
+      "acc": 0.8929534,
+      "epoch": 1.0072895957587806,
+      "grad_norm": 6.683059805886383,
+      "learning_rate": 9.999996368417133e-06,
+      "loss": 0.52333064,
+      "memory(GiB)": 33.01,
+      "step": 3040,
+      "train_speed(iter/s)": 0.18665
+    },
+    {
+      "acc": 0.89874592,
+      "epoch": 1.0089463220675945,
+      "grad_norm": 3.324861288632335,
+      "learning_rate": 9.999994530116223e-06,
+      "loss": 0.51245885,
+      "memory(GiB)": 33.01,
+      "step": 3045,
+      "train_speed(iter/s)": 0.186659
+    },
+    {
+      "acc": 0.91399202,
+      "epoch": 1.0106030483764081,
+      "grad_norm": 2.200372721059653,
+      "learning_rate": 9.99999231665216e-06,
+      "loss": 0.45101566,
+      "memory(GiB)": 33.01,
+      "step": 3050,
+      "train_speed(iter/s)": 0.186668
+    },
+    {
+      "acc": 0.90618773,
+      "epoch": 1.012259774685222,
+      "grad_norm": 3.197535643791647,
+      "learning_rate": 9.999989728025112e-06,
+      "loss": 0.470046,
+      "memory(GiB)": 33.01,
+      "step": 3055,
+      "train_speed(iter/s)": 0.186677
+    },
+    {
+      "acc": 0.89488354,
+      "epoch": 1.0139165009940359,
+      "grad_norm": 3.6610620112383843,
+      "learning_rate": 9.99998676423527e-06,
+      "loss": 0.49543657,
+      "memory(GiB)": 33.01,
+      "step": 3060,
+      "train_speed(iter/s)": 0.186686
+    },
+    {
+      "acc": 0.90282478,
+      "epoch": 1.0155732273028495,
+      "grad_norm": 7.12331538371565,
+      "learning_rate": 9.99998342528286e-06,
+      "loss": 0.47873211,
+      "memory(GiB)": 33.01,
+      "step": 3065,
+      "train_speed(iter/s)": 0.186696
+    },
+    {
+      "acc": 0.90743256,
+      "epoch": 1.0172299536116634,
+      "grad_norm": 7.3538832979952,
+      "learning_rate": 9.999979711168132e-06,
+      "loss": 0.48369231,
+      "memory(GiB)": 33.01,
+      "step": 3070,
+      "train_speed(iter/s)": 0.186705
+    },
+    {
+      "acc": 0.89925756,
+      "epoch": 1.0188866799204772,
+      "grad_norm": 5.3141924394185915,
+      "learning_rate": 9.999975621891365e-06,
+      "loss": 0.51017771,
+      "memory(GiB)": 33.01,
+      "step": 3075,
+      "train_speed(iter/s)": 0.186715
+    },
+    {
+      "acc": 0.89403591,
+      "epoch": 1.0205434062292909,
+      "grad_norm": 2.901880423400501,
+      "learning_rate": 9.999971157452864e-06,
+      "loss": 0.50029345,
+      "memory(GiB)": 33.01,
+      "step": 3080,
+      "train_speed(iter/s)": 0.186724
+    },
+    {
+      "acc": 0.89361725,
+      "epoch": 1.0222001325381047,
+      "grad_norm": 2.4117637892733326,
+      "learning_rate": 9.999966317852965e-06,
+      "loss": 0.52009544,
+      "memory(GiB)": 33.01,
+      "step": 3085,
+      "train_speed(iter/s)": 0.186718
+    },
+    {
+      "acc": 0.89835072,
+      "epoch": 1.0238568588469186,
+      "grad_norm": 2.750198930871617,
+      "learning_rate": 9.999961103092028e-06,
+      "loss": 0.52763042,
+      "memory(GiB)": 33.01,
+      "step": 3090,
+      "train_speed(iter/s)": 0.186711
+    },
+    {
+      "acc": 0.89915695,
+      "epoch": 1.0255135851557322,
+      "grad_norm": 3.95803224526299,
+      "learning_rate": 9.99995551317045e-06,
+      "loss": 0.51246595,
+      "memory(GiB)": 33.01,
+      "step": 3095,
+      "train_speed(iter/s)": 0.186716
+    },
+    {
+      "acc": 0.90447063,
+      "epoch": 1.027170311464546,
+      "grad_norm": 2.656686168940902,
+      "learning_rate": 9.999949548088648e-06,
+      "loss": 0.47282877,
+      "memory(GiB)": 33.01,
+      "step": 3100,
+      "train_speed(iter/s)": 0.186724
+    },
+    {
+      "acc": 0.9053257,
+      "epoch": 1.02882703777336,
+      "grad_norm": 2.702954123523646,
+      "learning_rate": 9.999943207847069e-06,
+      "loss": 0.45020781,
+      "memory(GiB)": 33.01,
+      "step": 3105,
+      "train_speed(iter/s)": 0.186734
+    },
+    {
+      "acc": 0.90436687,
+      "epoch": 1.0304837640821736,
+      "grad_norm": 3.518867930567591,
+      "learning_rate": 9.999936492446188e-06,
+      "loss": 0.46951027,
+      "memory(GiB)": 33.01,
+      "step": 3110,
+      "train_speed(iter/s)": 0.186743
+    },
+    {
+      "acc": 0.90386429,
+      "epoch": 1.0321404903909874,
+      "grad_norm": 3.2722257223182325,
+      "learning_rate": 9.999929401886511e-06,
+      "loss": 0.46613092,
+      "memory(GiB)": 33.01,
+      "step": 3115,
+      "train_speed(iter/s)": 0.186752
+    },
+    {
+      "acc": 0.90762558,
+      "epoch": 1.0337972166998013,
+      "grad_norm": 3.6064831547615412,
+      "learning_rate": 9.999921936168569e-06,
+      "loss": 0.47079697,
+      "memory(GiB)": 33.01,
+      "step": 3120,
+      "train_speed(iter/s)": 0.186761
+    },
+    {
+      "acc": 0.90466404,
+      "epoch": 1.035453943008615,
+      "grad_norm": 24.535623163011532,
+      "learning_rate": 9.999914095292922e-06,
+      "loss": 0.48182907,
+      "memory(GiB)": 33.01,
+      "step": 3125,
+      "train_speed(iter/s)": 0.186771
+    },
+    {
+      "acc": 0.90340786,
+      "epoch": 1.0371106693174288,
+      "grad_norm": 2.9159017707204695,
+      "learning_rate": 9.999905879260158e-06,
+      "loss": 0.46948576,
+      "memory(GiB)": 33.01,
+      "step": 3130,
+      "train_speed(iter/s)": 0.18678
+    },
+    {
+      "acc": 0.8990778,
+      "epoch": 1.0387673956262427,
+      "grad_norm": 3.190435152275673,
+      "learning_rate": 9.999897288070895e-06,
+      "loss": 0.53192158,
+      "memory(GiB)": 33.01,
+      "step": 3135,
+      "train_speed(iter/s)": 0.186789
+    },
+    {
+      "acc": 0.8878931,
+      "epoch": 1.0404241219350563,
+      "grad_norm": 3.984589063314788,
+      "learning_rate": 9.999888321725778e-06,
+      "loss": 0.54462519,
+      "memory(GiB)": 33.01,
+      "step": 3140,
+      "train_speed(iter/s)": 0.186798
+    },
+    {
+      "acc": 0.9036356,
+      "epoch": 1.0420808482438702,
+      "grad_norm": 4.1234824814065005,
+      "learning_rate": 9.999878980225477e-06,
+      "loss": 0.48756065,
+      "memory(GiB)": 33.01,
+      "step": 3145,
+      "train_speed(iter/s)": 0.186807
+    },
+    {
+      "acc": 0.90776148,
+      "epoch": 1.0437375745526838,
+      "grad_norm": 3.184455308667677,
+      "learning_rate": 9.999869263570696e-06,
+      "loss": 0.48718882,
+      "memory(GiB)": 33.01,
+      "step": 3150,
+      "train_speed(iter/s)": 0.186816
+    },
+    {
+      "acc": 0.90270214,
+      "epoch": 1.0453943008614976,
+      "grad_norm": 2.879851373753325,
+      "learning_rate": 9.999859171762164e-06,
+      "loss": 0.4768774,
+      "memory(GiB)": 33.01,
+      "step": 3155,
+      "train_speed(iter/s)": 0.186825
+    },
+    {
+      "acc": 0.9136734,
+      "epoch": 1.0470510271703115,
+      "grad_norm": 2.804970085971501,
+      "learning_rate": 9.999848704800635e-06,
+      "loss": 0.42542081,
+      "memory(GiB)": 33.01,
+      "step": 3160,
+      "train_speed(iter/s)": 0.186834
+    },
+    {
+      "acc": 0.90003986,
+      "epoch": 1.0487077534791251,
+      "grad_norm": 2.2480285866426275,
+      "learning_rate": 9.999837862686897e-06,
+      "loss": 0.46268835,
+      "memory(GiB)": 33.01,
+      "step": 3165,
+      "train_speed(iter/s)": 0.186843
+    },
+    {
+      "acc": 0.90802698,
+      "epoch": 1.050364479787939,
+      "grad_norm": 3.252978868283233,
+      "learning_rate": 9.999826645421765e-06,
+      "loss": 0.44581494,
+      "memory(GiB)": 33.01,
+      "step": 3170,
+      "train_speed(iter/s)": 0.186851
+    },
+    {
+      "acc": 0.89781418,
+      "epoch": 1.0520212060967529,
+      "grad_norm": 4.894761130520978,
+      "learning_rate": 9.999815053006078e-06,
+      "loss": 0.51909595,
+      "memory(GiB)": 33.01,
+      "step": 3175,
+      "train_speed(iter/s)": 0.18686
+    },
+    {
+      "acc": 0.89423304,
+      "epoch": 1.0536779324055665,
+      "grad_norm": 4.716050857292568,
+      "learning_rate": 9.999803085440708e-06,
+      "loss": 0.54763927,
+      "memory(GiB)": 33.01,
+      "step": 3180,
+      "train_speed(iter/s)": 0.186869
+    },
+    {
+      "acc": 0.90006123,
+      "epoch": 1.0553346587143804,
+      "grad_norm": 2.617206332092734,
+      "learning_rate": 9.99979074272655e-06,
+      "loss": 0.48034215,
+      "memory(GiB)": 33.01,
+      "step": 3185,
+      "train_speed(iter/s)": 0.186877
+    },
+    {
+      "acc": 0.89743977,
+      "epoch": 1.0569913850231942,
+      "grad_norm": 3.2918068791076225,
+      "learning_rate": 9.999778024864532e-06,
+      "loss": 0.50724802,
+      "memory(GiB)": 33.01,
+      "step": 3190,
+      "train_speed(iter/s)": 0.186886
+    },
+    {
+      "acc": 0.90672321,
+      "epoch": 1.0586481113320079,
+      "grad_norm": 3.5955126458372857,
+      "learning_rate": 9.999764931855611e-06,
+      "loss": 0.47439051,
+      "memory(GiB)": 33.01,
+      "step": 3195,
+      "train_speed(iter/s)": 0.186895
+    },
+    {
+      "acc": 0.89568729,
+      "epoch": 1.0603048376408217,
+      "grad_norm": 3.5230454219970184,
+      "learning_rate": 9.999751463700765e-06,
+      "loss": 0.4475791,
+      "memory(GiB)": 33.01,
+      "step": 3200,
+      "train_speed(iter/s)": 0.186904
+    },
+    {
+      "acc": 0.91135111,
+      "epoch": 1.0619615639496356,
+      "grad_norm": 2.190094029746288,
+      "learning_rate": 9.999737620401006e-06,
+      "loss": 0.41744318,
+      "memory(GiB)": 33.01,
+      "step": 3205,
+      "train_speed(iter/s)": 0.186913
+    },
+    {
+      "acc": 0.90460682,
+      "epoch": 1.0636182902584492,
+      "grad_norm": 3.3341024121046825,
+      "learning_rate": 9.999723401957376e-06,
+      "loss": 0.41109085,
+      "memory(GiB)": 33.01,
+      "step": 3210,
+      "train_speed(iter/s)": 0.186921
+    },
+    {
+      "acc": 0.90452099,
+      "epoch": 1.065275016567263,
+      "grad_norm": 4.172841742216287,
+      "learning_rate": 9.999708808370937e-06,
+      "loss": 0.45828161,
+      "memory(GiB)": 33.01,
+      "step": 3215,
+      "train_speed(iter/s)": 0.186929
+    },
+    {
+      "acc": 0.911588,
+      "epoch": 1.066931742876077,
+      "grad_norm": 3.6576288539358375,
+      "learning_rate": 9.999693839642787e-06,
+      "loss": 0.45138202,
+      "memory(GiB)": 33.01,
+      "step": 3220,
+      "train_speed(iter/s)": 0.186937
+    },
+    {
+      "acc": 0.90843134,
+      "epoch": 1.0685884691848906,
+      "grad_norm": 3.9526372468287194,
+      "learning_rate": 9.99967849577405e-06,
+      "loss": 0.42826166,
+      "memory(GiB)": 33.01,
+      "step": 3225,
+      "train_speed(iter/s)": 0.186946
+    },
+    {
+      "acc": 0.90101509,
+      "epoch": 1.0702451954937044,
+      "grad_norm": 4.050874644927262,
+      "learning_rate": 9.999662776765875e-06,
+      "loss": 0.47177067,
+      "memory(GiB)": 33.01,
+      "step": 3230,
+      "train_speed(iter/s)": 0.186954
+    },
+    {
+      "acc": 0.89944954,
+      "epoch": 1.0719019218025183,
+      "grad_norm": 3.5071629101682724,
+      "learning_rate": 9.999646682619443e-06,
+      "loss": 0.46376143,
+      "memory(GiB)": 33.01,
+      "step": 3235,
+      "train_speed(iter/s)": 0.186963
+    },
+    {
+      "acc": 0.90557594,
+      "epoch": 1.073558648111332,
+      "grad_norm": 3.2694798381027628,
+      "learning_rate": 9.999630213335963e-06,
+      "loss": 0.48373728,
+      "memory(GiB)": 33.01,
+      "step": 3240,
+      "train_speed(iter/s)": 0.186971
+    },
+    {
+      "acc": 0.8997365,
+      "epoch": 1.0752153744201458,
+      "grad_norm": 4.088897152250184,
+      "learning_rate": 9.999613368916666e-06,
+      "loss": 0.4708004,
+      "memory(GiB)": 33.01,
+      "step": 3245,
+      "train_speed(iter/s)": 0.18698
+    },
+    {
+      "acc": 0.90957518,
+      "epoch": 1.0768721007289597,
+      "grad_norm": 3.5456188254242926,
+      "learning_rate": 9.999596149362823e-06,
+      "loss": 0.41807528,
+      "memory(GiB)": 33.01,
+      "step": 3250,
+      "train_speed(iter/s)": 0.186988
+    },
+    {
+      "acc": 0.90725698,
+      "epoch": 1.0785288270377733,
+      "grad_norm": 3.2934084922845868,
+      "learning_rate": 9.99957855467572e-06,
+      "loss": 0.41615124,
+      "memory(GiB)": 33.01,
+      "step": 3255,
+      "train_speed(iter/s)": 0.186996
+    },
+    {
+      "acc": 0.91363049,
+      "epoch": 1.0801855533465872,
+      "grad_norm": 2.9929364538347074,
+      "learning_rate": 9.99956058485668e-06,
+      "loss": 0.39844606,
+      "memory(GiB)": 33.01,
+      "step": 3260,
+      "train_speed(iter/s)": 0.187005
+    },
+    {
+      "acc": 0.92446899,
+      "epoch": 1.081842279655401,
+      "grad_norm": 3.0013491142983875,
+      "learning_rate": 9.999542239907052e-06,
+      "loss": 0.38574693,
+      "memory(GiB)": 33.01,
+      "step": 3265,
+      "train_speed(iter/s)": 0.187013
+    },
+    {
+      "acc": 0.91798916,
+      "epoch": 1.0834990059642147,
+      "grad_norm": 2.9253649675762707,
+      "learning_rate": 9.999523519828212e-06,
+      "loss": 0.38218203,
+      "memory(GiB)": 33.01,
+      "step": 3270,
+      "train_speed(iter/s)": 0.187022
+    },
+    {
+      "acc": 0.91032162,
+      "epoch": 1.0851557322730285,
+      "grad_norm": 3.4505921508829287,
+      "learning_rate": 9.999504424621564e-06,
+      "loss": 0.40374975,
+      "memory(GiB)": 33.01,
+      "step": 3275,
+      "train_speed(iter/s)": 0.187029
+    },
+    {
+      "acc": 0.90089769,
+      "epoch": 1.0868124585818424,
+      "grad_norm": 3.6954550013229697,
+      "learning_rate": 9.999484954288543e-06,
+      "loss": 0.47235622,
+      "memory(GiB)": 33.01,
+      "step": 3280,
+      "train_speed(iter/s)": 0.187037
+    },
+    {
+      "acc": 0.90761738,
+      "epoch": 1.088469184890656,
+      "grad_norm": 3.5433394473860798,
+      "learning_rate": 9.999465108830605e-06,
+      "loss": 0.41755667,
+      "memory(GiB)": 33.01,
+      "step": 3285,
+      "train_speed(iter/s)": 0.187046
+    },
+    {
+      "acc": 0.91872139,
+      "epoch": 1.0901259111994699,
+      "grad_norm": 3.9994350312871383,
+      "learning_rate": 9.999444888249247e-06,
+      "loss": 0.43578491,
+      "memory(GiB)": 33.01,
+      "step": 3290,
+      "train_speed(iter/s)": 0.187054
+    },
+    {
+      "acc": 0.92029285,
+      "epoch": 1.0917826375082837,
+      "grad_norm": 2.8535836540855994,
+      "learning_rate": 9.999424292545981e-06,
+      "loss": 0.40754728,
+      "memory(GiB)": 33.01,
+      "step": 3295,
+      "train_speed(iter/s)": 0.187063
+    },
+    {
+      "acc": 0.91320772,
+      "epoch": 1.0934393638170974,
+      "grad_norm": 3.159190628284286,
+      "learning_rate": 9.999403321722352e-06,
+      "loss": 0.44989252,
+      "memory(GiB)": 33.01,
+      "step": 3300,
+      "train_speed(iter/s)": 0.187072
+    },
+    {
+      "acc": 0.90615292,
+      "epoch": 1.0950960901259112,
+      "grad_norm": 2.6560869687241637,
+      "learning_rate": 9.999381975779937e-06,
+      "loss": 0.41778717,
+      "memory(GiB)": 33.01,
+      "step": 3305,
+      "train_speed(iter/s)": 0.18708
+    },
+    {
+      "acc": 0.91706696,
+      "epoch": 1.0967528164347249,
+      "grad_norm": 3.3806554149021606,
+      "learning_rate": 9.999360254720336e-06,
+      "loss": 0.40516958,
+      "memory(GiB)": 33.01,
+      "step": 3310,
+      "train_speed(iter/s)": 0.187082
+    },
+    {
+      "acc": 0.91830273,
+      "epoch": 1.0984095427435387,
+      "grad_norm": 3.019253191194386,
+      "learning_rate": 9.999338158545177e-06,
+      "loss": 0.39181962,
+      "memory(GiB)": 33.01,
+      "step": 3315,
+      "train_speed(iter/s)": 0.187083
+    },
+    {
+      "acc": 0.91280575,
+      "epoch": 1.1000662690523526,
+      "grad_norm": 3.086005010991861,
+      "learning_rate": 9.999315687256122e-06,
+      "loss": 0.41425366,
+      "memory(GiB)": 33.01,
+      "step": 3320,
+      "train_speed(iter/s)": 0.187085
+    },
+    {
+      "acc": 0.91823215,
+      "epoch": 1.1017229953611662,
+      "grad_norm": 2.5142824427576036,
+      "learning_rate": 9.999292840854854e-06,
+      "loss": 0.36637301,
+      "memory(GiB)": 33.01,
+      "step": 3325,
+      "train_speed(iter/s)": 0.187089
+    },
+    {
+      "acc": 0.91018171,
+      "epoch": 1.10337972166998,
+      "grad_norm": 4.8237007090091915,
+      "learning_rate": 9.99926961934309e-06,
+      "loss": 0.43959985,
+      "memory(GiB)": 33.01,
+      "step": 3330,
+      "train_speed(iter/s)": 0.187098
+    },
+    {
+      "acc": 0.91465988,
+      "epoch": 1.105036447978794,
+      "grad_norm": 5.9902259481845554,
+      "learning_rate": 9.999246022722572e-06,
+      "loss": 0.42311144,
+      "memory(GiB)": 33.01,
+      "step": 3335,
+      "train_speed(iter/s)": 0.187107
+    },
+    {
+      "acc": 0.90747385,
+      "epoch": 1.1066931742876076,
+      "grad_norm": 3.742504529402083,
+      "learning_rate": 9.999222050995069e-06,
+      "loss": 0.41870995,
+      "memory(GiB)": 33.01,
+      "step": 3340,
+      "train_speed(iter/s)": 0.187115
+    },
+    {
+      "acc": 0.9087347,
+      "epoch": 1.1083499005964215,
+      "grad_norm": 3.081075463249609,
+      "learning_rate": 9.999197704162381e-06,
+      "loss": 0.43444052,
+      "memory(GiB)": 33.01,
+      "step": 3345,
+      "train_speed(iter/s)": 0.187123
+    },
+    {
+      "acc": 0.9046854,
+      "epoch": 1.1100066269052353,
+      "grad_norm": 4.796168628118214,
+      "learning_rate": 9.999172982226335e-06,
+      "loss": 0.47512655,
+      "memory(GiB)": 33.01,
+      "step": 3350,
+      "train_speed(iter/s)": 0.187132
+    },
+    {
+      "acc": 0.90798407,
+      "epoch": 1.111663353214049,
+      "grad_norm": 2.98283344383208,
+      "learning_rate": 9.999147885188786e-06,
+      "loss": 0.44967909,
+      "memory(GiB)": 33.01,
+      "step": 3355,
+      "train_speed(iter/s)": 0.18714
+    },
+    {
+      "acc": 0.9132597,
+      "epoch": 1.1133200795228628,
+      "grad_norm": 5.148743019636876,
+      "learning_rate": 9.999122413051616e-06,
+      "loss": 0.41550369,
+      "memory(GiB)": 33.01,
+      "step": 3360,
+      "train_speed(iter/s)": 0.187148
+    },
+    {
+      "acc": 0.91174011,
+      "epoch": 1.1149768058316767,
+      "grad_norm": 4.429328627712513,
+      "learning_rate": 9.999096565816741e-06,
+      "loss": 0.40560884,
+      "memory(GiB)": 33.01,
+      "step": 3365,
+      "train_speed(iter/s)": 0.187156
+    },
+    {
+      "acc": 0.9127593,
+      "epoch": 1.1166335321404903,
+      "grad_norm": 2.7428037575839994,
+      "learning_rate": 9.999070343486095e-06,
+      "loss": 0.38106241,
+      "memory(GiB)": 33.01,
+      "step": 3370,
+      "train_speed(iter/s)": 0.187164
+    },
+    {
+      "acc": 0.91239586,
+      "epoch": 1.1182902584493042,
+      "grad_norm": 3.6801501583797127,
+      "learning_rate": 9.999043746061649e-06,
+      "loss": 0.39418139,
+      "memory(GiB)": 33.01,
+      "step": 3375,
+      "train_speed(iter/s)": 0.187172
+    },
+    {
+      "acc": 0.91507854,
+      "epoch": 1.119946984758118,
+      "grad_norm": 3.6826512051892926,
+      "learning_rate": 9.999016773545396e-06,
+      "loss": 0.40172067,
+      "memory(GiB)": 33.01,
+      "step": 3380,
+      "train_speed(iter/s)": 0.18718
+    },
+    {
+      "acc": 0.91697407,
+      "epoch": 1.1216037110669317,
+      "grad_norm": 3.7316676664290243,
+      "learning_rate": 9.998989425939364e-06,
+      "loss": 0.38764234,
+      "memory(GiB)": 33.01,
+      "step": 3385,
+      "train_speed(iter/s)": 0.187188
+    },
+    {
+      "acc": 0.90916672,
+      "epoch": 1.1232604373757455,
+      "grad_norm": 3.118760160636055,
+      "learning_rate": 9.998961703245603e-06,
+      "loss": 0.40729375,
+      "memory(GiB)": 33.01,
+      "step": 3390,
+      "train_speed(iter/s)": 0.187196
+    },
+    {
+      "acc": 0.91550961,
+      "epoch": 1.1249171636845594,
+      "grad_norm": 3.402058127498593,
+      "learning_rate": 9.998933605466193e-06,
+      "loss": 0.36972272,
+      "memory(GiB)": 33.01,
+      "step": 3395,
+      "train_speed(iter/s)": 0.187204
+    },
+    {
+      "acc": 0.92313728,
+      "epoch": 1.126573889993373,
+      "grad_norm": 3.2158022961513244,
+      "learning_rate": 9.998905132603244e-06,
+      "loss": 0.40722723,
+      "memory(GiB)": 33.01,
+      "step": 3400,
+      "train_speed(iter/s)": 0.187212
+    },
+    {
+      "acc": 0.91641541,
+      "epoch": 1.128230616302187,
+      "grad_norm": 3.4663386627981265,
+      "learning_rate": 9.99887628465889e-06,
+      "loss": 0.36683979,
+      "memory(GiB)": 33.01,
+      "step": 3405,
+      "train_speed(iter/s)": 0.18722
+    },
+    {
+      "acc": 0.90837517,
+      "epoch": 1.1298873426110005,
+      "grad_norm": 3.406324627196776,
+      "learning_rate": 9.998847061635298e-06,
+      "loss": 0.45863819,
+      "memory(GiB)": 33.01,
+      "step": 3410,
+      "train_speed(iter/s)": 0.187229
+    },
+    {
+      "acc": 0.90772152,
+      "epoch": 1.1315440689198144,
+      "grad_norm": 3.5219354346318,
+      "learning_rate": 9.998817463534658e-06,
+      "loss": 0.43631525,
+      "memory(GiB)": 33.01,
+      "step": 3415,
+      "train_speed(iter/s)": 0.187237
+    },
+    {
+      "acc": 0.90815153,
+      "epoch": 1.1332007952286283,
+      "grad_norm": 1.8801903029118625,
+      "learning_rate": 9.998787490359197e-06,
+      "loss": 0.39505372,
+      "memory(GiB)": 33.01,
+      "step": 3420,
+      "train_speed(iter/s)": 0.187245
+    },
+    {
+      "acc": 0.92017059,
+      "epoch": 1.134857521537442,
+      "grad_norm": 3.439583077856211,
+      "learning_rate": 9.998757142111157e-06,
+      "loss": 0.36194487,
+      "memory(GiB)": 33.01,
+      "step": 3425,
+      "train_speed(iter/s)": 0.187253
+    },
+    {
+      "acc": 0.91789627,
+      "epoch": 1.1365142478462558,
+      "grad_norm": 3.3683622265052215,
+      "learning_rate": 9.99872641879282e-06,
+      "loss": 0.37170382,
+      "memory(GiB)": 33.01,
+      "step": 3430,
+      "train_speed(iter/s)": 0.18726
+    },
+    {
+      "acc": 0.91388178,
+      "epoch": 1.1381709741550696,
+      "grad_norm": 3.317847183719663,
+      "learning_rate": 9.998695320406492e-06,
+      "loss": 0.40776911,
+      "memory(GiB)": 33.01,
+      "step": 3435,
+      "train_speed(iter/s)": 0.187268
+    },
+    {
+      "acc": 0.90797386,
+      "epoch": 1.1398277004638833,
+      "grad_norm": 2.6200821673604957,
+      "learning_rate": 9.998663846954502e-06,
+      "loss": 0.39878166,
+      "memory(GiB)": 33.01,
+      "step": 3440,
+      "train_speed(iter/s)": 0.187276
+    },
+    {
+      "acc": 0.91855488,
+      "epoch": 1.1414844267726971,
+      "grad_norm": 2.8307763205024603,
+      "learning_rate": 9.998631998439218e-06,
+      "loss": 0.40054965,
+      "memory(GiB)": 33.01,
+      "step": 3445,
+      "train_speed(iter/s)": 0.187284
+    },
+    {
+      "acc": 0.91904821,
+      "epoch": 1.143141153081511,
+      "grad_norm": 3.696799291377706,
+      "learning_rate": 9.998599774863023e-06,
+      "loss": 0.38020282,
+      "memory(GiB)": 33.01,
+      "step": 3450,
+      "train_speed(iter/s)": 0.187291
+    },
+    {
+      "acc": 0.9194087,
+      "epoch": 1.1447978793903246,
+      "grad_norm": 3.2008164309337364,
+      "learning_rate": 9.998567176228341e-06,
+      "loss": 0.38544326,
+      "memory(GiB)": 33.01,
+      "step": 3455,
+      "train_speed(iter/s)": 0.187299
+    },
+    {
+      "acc": 0.91780272,
+      "epoch": 1.1464546056991385,
+      "grad_norm": 2.840912562079474,
+      "learning_rate": 9.998534202537615e-06,
+      "loss": 0.37580428,
+      "memory(GiB)": 33.01,
+      "step": 3460,
+      "train_speed(iter/s)": 0.187307
+    },
+    {
+      "acc": 0.91689548,
+      "epoch": 1.1481113320079523,
+      "grad_norm": 3.851478472278105,
+      "learning_rate": 9.99850085379332e-06,
+      "loss": 0.40554276,
+      "memory(GiB)": 33.01,
+      "step": 3465,
+      "train_speed(iter/s)": 0.187314
+    },
+    {
+      "acc": 0.91284361,
+      "epoch": 1.149768058316766,
+      "grad_norm": 3.2849355891803858,
+      "learning_rate": 9.99846712999796e-06,
+      "loss": 0.38951898,
+      "memory(GiB)": 33.01,
+      "step": 3470,
+      "train_speed(iter/s)": 0.187322
+    },
+    {
+      "acc": 0.90839863,
+      "epoch": 1.1514247846255798,
+      "grad_norm": 3.633392809645267,
+      "learning_rate": 9.998433031154064e-06,
+      "loss": 0.39964263,
+      "memory(GiB)": 33.01,
+      "step": 3475,
+      "train_speed(iter/s)": 0.187329
+    },
+    {
+      "acc": 0.92207413,
+      "epoch": 1.1530815109343937,
+      "grad_norm": 2.5154870578677477,
+      "learning_rate": 9.998398557264191e-06,
+      "loss": 0.37701032,
+      "memory(GiB)": 33.01,
+      "step": 3480,
+      "train_speed(iter/s)": 0.187337
+    },
+    {
+      "acc": 0.91332684,
+      "epoch": 1.1547382372432073,
+      "grad_norm": 3.25480052804041,
+      "learning_rate": 9.998363708330927e-06,
+      "loss": 0.40475931,
+      "memory(GiB)": 33.01,
+      "step": 3485,
+      "train_speed(iter/s)": 0.187344
+    },
+    {
+      "acc": 0.9196701,
+      "epoch": 1.1563949635520212,
+      "grad_norm": 2.3516656649413874,
+      "learning_rate": 9.998328484356889e-06,
+      "loss": 0.39389837,
+      "memory(GiB)": 33.01,
+      "step": 3490,
+      "train_speed(iter/s)": 0.187352
+    },
+    {
+      "acc": 0.92585783,
+      "epoch": 1.158051689860835,
+      "grad_norm": 2.8555420991231295,
+      "learning_rate": 9.998292885344719e-06,
+      "loss": 0.37741697,
+      "memory(GiB)": 33.01,
+      "step": 3495,
+      "train_speed(iter/s)": 0.18736
+    },
+    {
+      "acc": 0.91590786,
+      "epoch": 1.1597084161696487,
+      "grad_norm": 2.5764556931798053,
+      "learning_rate": 9.99825691129709e-06,
+      "loss": 0.36057463,
+      "memory(GiB)": 33.01,
+      "step": 3500,
+      "train_speed(iter/s)": 0.187368
+    },
+    {
+      "acc": 0.92557907,
+      "epoch": 1.1613651424784626,
+      "grad_norm": 2.5339370927921654,
+      "learning_rate": 9.998220562216697e-06,
+      "loss": 0.37259173,
+      "memory(GiB)": 33.01,
+      "step": 3505,
+      "train_speed(iter/s)": 0.187376
+    },
+    {
+      "acc": 0.92664833,
+      "epoch": 1.1630218687872764,
+      "grad_norm": 2.273777795908151,
+      "learning_rate": 9.998183838106274e-06,
+      "loss": 0.33274937,
+      "memory(GiB)": 33.01,
+      "step": 3510,
+      "train_speed(iter/s)": 0.187383
+    },
+    {
+      "acc": 0.92987843,
+      "epoch": 1.16467859509609,
+      "grad_norm": 4.244783691017268,
+      "learning_rate": 9.998146738968571e-06,
+      "loss": 0.34606485,
+      "memory(GiB)": 33.01,
+      "step": 3515,
+      "train_speed(iter/s)": 0.187391
+    },
+    {
+      "acc": 0.92926168,
+      "epoch": 1.166335321404904,
+      "grad_norm": 2.886560446164786,
+      "learning_rate": 9.998109264806377e-06,
+      "loss": 0.36356709,
+      "memory(GiB)": 33.01,
+      "step": 3520,
+      "train_speed(iter/s)": 0.187398
+    },
+    {
+      "acc": 0.92352371,
+      "epoch": 1.1679920477137178,
+      "grad_norm": 3.5886682050868384,
+      "learning_rate": 9.9980714156225e-06,
+      "loss": 0.36706614,
+      "memory(GiB)": 33.01,
+      "step": 3525,
+      "train_speed(iter/s)": 0.187406
+    },
+    {
+      "acc": 0.92344646,
+      "epoch": 1.1696487740225314,
+      "grad_norm": 3.1139911670284848,
+      "learning_rate": 9.998033191419783e-06,
+      "loss": 0.33353124,
+      "memory(GiB)": 33.01,
+      "step": 3530,
+      "train_speed(iter/s)": 0.187414
+    },
+    {
+      "acc": 0.92489176,
+      "epoch": 1.1713055003313453,
+      "grad_norm": 3.2696372919157977,
+      "learning_rate": 9.997994592201091e-06,
+      "loss": 0.36244659,
+      "memory(GiB)": 33.01,
+      "step": 3535,
+      "train_speed(iter/s)": 0.187422
+    },
+    {
+      "acc": 0.92667379,
+      "epoch": 1.1729622266401591,
+      "grad_norm": 2.840594409063846,
+      "learning_rate": 9.997955617969324e-06,
+      "loss": 0.33494053,
+      "memory(GiB)": 33.01,
+      "step": 3540,
+      "train_speed(iter/s)": 0.187422
+    },
+    {
+      "acc": 0.93094053,
+      "epoch": 1.1746189529489728,
+      "grad_norm": 2.6007626989163803,
+      "learning_rate": 9.997916268727406e-06,
+      "loss": 0.32520399,
+      "memory(GiB)": 33.01,
+      "step": 3545,
+      "train_speed(iter/s)": 0.187424
+    },
+    {
+      "acc": 0.92882347,
+      "epoch": 1.1762756792577866,
+      "grad_norm": 2.1525868787398847,
+      "learning_rate": 9.997876544478288e-06,
+      "loss": 0.34395452,
+      "memory(GiB)": 33.01,
+      "step": 3550,
+      "train_speed(iter/s)": 0.187422
+    },
+    {
+      "acc": 0.91319542,
+      "epoch": 1.1779324055666005,
+      "grad_norm": 4.288583356910695,
+      "learning_rate": 9.997836445224954e-06,
+      "loss": 0.37349844,
+      "memory(GiB)": 33.01,
+      "step": 3555,
+      "train_speed(iter/s)": 0.18743
+    },
+    {
+      "acc": 0.90995693,
+      "epoch": 1.1795891318754141,
+      "grad_norm": 2.533392777088841,
+      "learning_rate": 9.997795970970409e-06,
+      "loss": 0.39516404,
+      "memory(GiB)": 33.01,
+      "step": 3560,
+      "train_speed(iter/s)": 0.187438
+    },
+    {
+      "acc": 0.91162281,
+      "epoch": 1.181245858184228,
+      "grad_norm": 3.4627809496510094,
+      "learning_rate": 9.997755121717694e-06,
+      "loss": 0.43464499,
+      "memory(GiB)": 33.01,
+      "step": 3565,
+      "train_speed(iter/s)": 0.187446
+    },
+    {
+      "acc": 0.91588955,
+      "epoch": 1.1829025844930419,
+      "grad_norm": 4.111033674930325,
+      "learning_rate": 9.997713897469871e-06,
+      "loss": 0.38906169,
+      "memory(GiB)": 33.01,
+      "step": 3570,
+      "train_speed(iter/s)": 0.187453
+    },
+    {
+      "acc": 0.91391239,
+      "epoch": 1.1845593108018555,
+      "grad_norm": 3.1895353282084073,
+      "learning_rate": 9.997672298230037e-06,
+      "loss": 0.39768865,
+      "memory(GiB)": 33.01,
+      "step": 3575,
+      "train_speed(iter/s)": 0.187461
+    },
+    {
+      "acc": 0.91793308,
+      "epoch": 1.1862160371106694,
+      "grad_norm": 2.7817667892630813,
+      "learning_rate": 9.99763032400131e-06,
+      "loss": 0.39883778,
+      "memory(GiB)": 33.01,
+      "step": 3580,
+      "train_speed(iter/s)": 0.187469
+    },
+    {
+      "acc": 0.92342186,
+      "epoch": 1.1878727634194832,
+      "grad_norm": 2.128217928171914,
+      "learning_rate": 9.99758797478684e-06,
+      "loss": 0.3766788,
+      "memory(GiB)": 33.01,
+      "step": 3585,
+      "train_speed(iter/s)": 0.187476
+    },
+    {
+      "acc": 0.9227787,
+      "epoch": 1.1895294897282969,
+      "grad_norm": 2.417063985621045,
+      "learning_rate": 9.99754525058981e-06,
+      "loss": 0.33557951,
+      "memory(GiB)": 33.01,
+      "step": 3590,
+      "train_speed(iter/s)": 0.187484
+    },
+    {
+      "acc": 0.92011023,
+      "epoch": 1.1911862160371107,
+      "grad_norm": 2.6827761651541944,
+      "learning_rate": 9.99750215141342e-06,
+      "loss": 0.35906065,
+      "memory(GiB)": 33.01,
+      "step": 3595,
+      "train_speed(iter/s)": 0.187491
+    },
+    {
+      "acc": 0.92309437,
+      "epoch": 1.1928429423459244,
+      "grad_norm": 3.2842965392742225,
+      "learning_rate": 9.997458677260906e-06,
+      "loss": 0.34788375,
+      "memory(GiB)": 33.01,
+      "step": 3600,
+      "train_speed(iter/s)": 0.187499
+    },
+    {
+      "acc": 0.92133265,
+      "epoch": 1.1944996686547382,
+      "grad_norm": 3.463747502790954,
+      "learning_rate": 9.997414828135532e-06,
+      "loss": 0.37525055,
+      "memory(GiB)": 33.01,
+      "step": 3605,
+      "train_speed(iter/s)": 0.187506
+    },
+    {
+      "acc": 0.93312187,
+      "epoch": 1.196156394963552,
+      "grad_norm": 4.264407587151082,
+      "learning_rate": 9.997370604040587e-06,
+      "loss": 0.31689456,
+      "memory(GiB)": 33.01,
+      "step": 3610,
+      "train_speed(iter/s)": 0.187514
+    },
+    {
+      "acc": 0.93518686,
+      "epoch": 1.1978131212723657,
+      "grad_norm": 3.0239619820220542,
+      "learning_rate": 9.997326004979391e-06,
+      "loss": 0.33057671,
+      "memory(GiB)": 33.01,
+      "step": 3615,
+      "train_speed(iter/s)": 0.187522
+    },
+    {
+      "acc": 0.91941586,
+      "epoch": 1.1994698475811796,
+      "grad_norm": 2.587705719863201,
+      "learning_rate": 9.997281030955288e-06,
+      "loss": 0.36802444,
+      "memory(GiB)": 33.01,
+      "step": 3620,
+      "train_speed(iter/s)": 0.187529
+    },
+    {
+      "acc": 0.91424217,
+      "epoch": 1.2011265738899934,
+      "grad_norm": 4.536591989811321,
+      "learning_rate": 9.997235681971656e-06,
+      "loss": 0.40987997,
+      "memory(GiB)": 33.01,
+      "step": 3625,
+      "train_speed(iter/s)": 0.187536
+    },
+    {
+      "acc": 0.93441477,
+      "epoch": 1.202783300198807,
+      "grad_norm": 3.8329063732002293,
+      "learning_rate": 9.997189958031895e-06,
+      "loss": 0.31681237,
+      "memory(GiB)": 33.01,
+      "step": 3630,
+      "train_speed(iter/s)": 0.187543
+    },
+    {
+      "acc": 0.9128912,
+      "epoch": 1.204440026507621,
+      "grad_norm": 3.702790490407351,
+      "learning_rate": 9.997143859139439e-06,
+      "loss": 0.3997297,
+      "memory(GiB)": 33.01,
+      "step": 3635,
+      "train_speed(iter/s)": 0.18755
+    },
+    {
+      "acc": 0.91567707,
+      "epoch": 1.2060967528164348,
+      "grad_norm": 3.384491618727864,
+      "learning_rate": 9.997097385297744e-06,
+      "loss": 0.38490405,
+      "memory(GiB)": 33.01,
+      "step": 3640,
+      "train_speed(iter/s)": 0.187558
+    },
+    {
+      "acc": 0.93141842,
+      "epoch": 1.2077534791252484,
+      "grad_norm": 2.914997162442394,
+      "learning_rate": 9.9970505365103e-06,
+      "loss": 0.34473834,
+      "memory(GiB)": 33.01,
+      "step": 3645,
+      "train_speed(iter/s)": 0.187565
+    },
+    {
+      "acc": 0.92478352,
+      "epoch": 1.2094102054340623,
+      "grad_norm": 2.3528754980819504,
+      "learning_rate": 9.997003312780622e-06,
+      "loss": 0.35598297,
+      "memory(GiB)": 33.01,
+      "step": 3650,
+      "train_speed(iter/s)": 0.187573
+    },
+    {
+      "acc": 0.92494583,
+      "epoch": 1.2110669317428762,
+      "grad_norm": 6.933560996174331,
+      "learning_rate": 9.996955714112254e-06,
+      "loss": 0.35216231,
+      "memory(GiB)": 33.01,
+      "step": 3655,
+      "train_speed(iter/s)": 0.18758
+    },
+    {
+      "acc": 0.9253911,
+      "epoch": 1.2127236580516898,
+      "grad_norm": 3.5739708668035495,
+      "learning_rate": 9.996907740508767e-06,
+      "loss": 0.35801926,
+      "memory(GiB)": 33.01,
+      "step": 3660,
+      "train_speed(iter/s)": 0.187587
+    },
+    {
+      "acc": 0.92626677,
+      "epoch": 1.2143803843605037,
+      "grad_norm": 4.233605139073024,
+      "learning_rate": 9.996859391973761e-06,
+      "loss": 0.38392704,
+      "memory(GiB)": 33.01,
+      "step": 3665,
+      "train_speed(iter/s)": 0.187594
+    },
+    {
+      "acc": 0.92703028,
+      "epoch": 1.2160371106693175,
+      "grad_norm": 3.634820256417589,
+      "learning_rate": 9.996810668510865e-06,
+      "loss": 0.34094188,
+      "memory(GiB)": 33.01,
+      "step": 3670,
+      "train_speed(iter/s)": 0.187601
+    },
+    {
+      "acc": 0.93061657,
+      "epoch": 1.2176938369781312,
+      "grad_norm": 5.114858856367563,
+      "learning_rate": 9.996761570123734e-06,
+      "loss": 0.31726093,
+      "memory(GiB)": 33.01,
+      "step": 3675,
+      "train_speed(iter/s)": 0.187608
+    },
+    {
+      "acc": 0.92885113,
+      "epoch": 1.219350563286945,
+      "grad_norm": 4.652957243697469,
+      "learning_rate": 9.996712096816054e-06,
+      "loss": 0.32563906,
+      "memory(GiB)": 33.01,
+      "step": 3680,
+      "train_speed(iter/s)": 0.187616
+    },
+    {
+      "acc": 0.9350296,
+      "epoch": 1.2210072895957587,
+      "grad_norm": 3.5677819539479896,
+      "learning_rate": 9.996662248591534e-06,
+      "loss": 0.31505063,
+      "memory(GiB)": 33.01,
+      "step": 3685,
+      "train_speed(iter/s)": 0.187623
+    },
+    {
+      "acc": 0.92970486,
+      "epoch": 1.2226640159045725,
+      "grad_norm": 3.7798542926882677,
+      "learning_rate": 9.996612025453919e-06,
+      "loss": 0.34854329,
+      "memory(GiB)": 33.01,
+      "step": 3690,
+      "train_speed(iter/s)": 0.18763
+    },
+    {
+      "acc": 0.92640114,
+      "epoch": 1.2243207422133864,
+      "grad_norm": 3.2775733400719997,
+      "learning_rate": 9.996561427406974e-06,
+      "loss": 0.34672508,
+      "memory(GiB)": 33.01,
+      "step": 3695,
+      "train_speed(iter/s)": 0.187637
+    },
+    {
+      "acc": 0.92301512,
+      "epoch": 1.2259774685222,
+      "grad_norm": 3.7070009363339094,
+      "learning_rate": 9.996510454454502e-06,
+      "loss": 0.33321366,
+      "memory(GiB)": 33.01,
+      "step": 3700,
+      "train_speed(iter/s)": 0.187644
+    },
+    {
+      "acc": 0.91633167,
+      "epoch": 1.2276341948310139,
+      "grad_norm": 3.9089476682504403,
+      "learning_rate": 9.996459106600319e-06,
+      "loss": 0.36949565,
+      "memory(GiB)": 33.01,
+      "step": 3705,
+      "train_speed(iter/s)": 0.187651
+    },
+    {
+      "acc": 0.92749081,
+      "epoch": 1.2292909211398277,
+      "grad_norm": 2.383069253984934,
+      "learning_rate": 9.996407383848283e-06,
+      "loss": 0.36472716,
+      "memory(GiB)": 33.01,
+      "step": 3710,
+      "train_speed(iter/s)": 0.187658
+    },
+    {
+      "acc": 0.92329865,
+      "epoch": 1.2309476474486414,
+      "grad_norm": 2.355621651870094,
+      "learning_rate": 9.996355286202277e-06,
+      "loss": 0.33665237,
+      "memory(GiB)": 33.01,
+      "step": 3715,
+      "train_speed(iter/s)": 0.187665
+    },
+    {
+      "acc": 0.92338848,
+      "epoch": 1.2326043737574552,
+      "grad_norm": 2.982481024537641,
+      "learning_rate": 9.996302813666209e-06,
+      "loss": 0.35028746,
+      "memory(GiB)": 33.01,
+      "step": 3720,
+      "train_speed(iter/s)": 0.187672
+    },
+    {
+      "acc": 0.92760925,
+      "epoch": 1.234261100066269,
+      "grad_norm": 2.2523268776096006,
+      "learning_rate": 9.996249966244015e-06,
+      "loss": 0.30640817,
+      "memory(GiB)": 33.01,
+      "step": 3725,
+      "train_speed(iter/s)": 0.187679
+    },
+    {
+      "acc": 0.91822443,
+      "epoch": 1.2359178263750827,
+      "grad_norm": 4.214879626463359,
+      "learning_rate": 9.996196743939663e-06,
+      "loss": 0.36053338,
+      "memory(GiB)": 33.01,
+      "step": 3730,
+      "train_speed(iter/s)": 0.187686
+    },
+    {
+      "acc": 0.9352869,
+      "epoch": 1.2375745526838966,
+      "grad_norm": 2.1221716411892793,
+      "learning_rate": 9.996143146757144e-06,
+      "loss": 0.31079593,
+      "memory(GiB)": 33.01,
+      "step": 3735,
+      "train_speed(iter/s)": 0.187692
+    },
+    {
+      "acc": 0.93244896,
+      "epoch": 1.2392312789927105,
+      "grad_norm": 2.7792772570584483,
+      "learning_rate": 9.99608917470048e-06,
+      "loss": 0.31752284,
+      "memory(GiB)": 33.01,
+      "step": 3740,
+      "train_speed(iter/s)": 0.187699
+    },
+    {
+      "acc": 0.92920828,
+      "epoch": 1.240888005301524,
+      "grad_norm": 4.50665904388833,
+      "learning_rate": 9.996034827773726e-06,
+      "loss": 0.31363287,
+      "memory(GiB)": 33.01,
+      "step": 3745,
+      "train_speed(iter/s)": 0.187706
+    },
+    {
+      "acc": 0.93048716,
+      "epoch": 1.242544731610338,
+      "grad_norm": 5.150303884251782,
+      "learning_rate": 9.995980105980956e-06,
+      "loss": 0.32176299,
+      "memory(GiB)": 33.01,
+      "step": 3750,
+      "train_speed(iter/s)": 0.187713
+    },
+    {
+      "acc": 0.92537174,
+      "epoch": 1.2442014579191518,
+      "grad_norm": 3.820148408737309,
+      "learning_rate": 9.995925009326277e-06,
+      "loss": 0.35507479,
+      "memory(GiB)": 33.01,
+      "step": 3755,
+      "train_speed(iter/s)": 0.18772
+    },
+    {
+      "acc": 0.92827415,
+      "epoch": 1.2458581842279655,
+      "grad_norm": 3.1128400564399414,
+      "learning_rate": 9.995869537813824e-06,
+      "loss": 0.36116066,
+      "memory(GiB)": 33.01,
+      "step": 3760,
+      "train_speed(iter/s)": 0.187727
+    },
+    {
+      "acc": 0.92336235,
+      "epoch": 1.2475149105367793,
+      "grad_norm": 3.080411070133688,
+      "learning_rate": 9.995813691447756e-06,
+      "loss": 0.3824018,
+      "memory(GiB)": 33.01,
+      "step": 3765,
+      "train_speed(iter/s)": 0.187728
+    },
+    {
+      "acc": 0.92660742,
+      "epoch": 1.2491716368455932,
+      "grad_norm": 2.616635669642493,
+      "learning_rate": 9.995757470232271e-06,
+      "loss": 0.3477334,
+      "memory(GiB)": 33.01,
+      "step": 3770,
+      "train_speed(iter/s)": 0.187729
+    },
+    {
+      "acc": 0.9220067,
+      "epoch": 1.2508283631544068,
+      "grad_norm": 3.102245923679212,
+      "learning_rate": 9.995700874171583e-06,
+      "loss": 0.3722064,
+      "memory(GiB)": 33.01,
+      "step": 3775,
+      "train_speed(iter/s)": 0.187728
+    },
+    {
+      "acc": 0.92387257,
+      "epoch": 1.2524850894632207,
+      "grad_norm": 3.4960303388785436,
+      "learning_rate": 9.995643903269939e-06,
+      "loss": 0.34432602,
+      "memory(GiB)": 33.01,
+      "step": 3780,
+      "train_speed(iter/s)": 0.187732
+    },
+    {
+      "acc": 0.92673817,
+      "epoch": 1.2541418157720345,
+      "grad_norm": 3.726521024830549,
+      "learning_rate": 9.995586557531614e-06,
+      "loss": 0.35871067,
+      "memory(GiB)": 33.01,
+      "step": 3785,
+      "train_speed(iter/s)": 0.187739
+    },
+    {
+      "acc": 0.92667179,
+      "epoch": 1.2557985420808482,
+      "grad_norm": 4.925958542452472,
+      "learning_rate": 9.995528836960913e-06,
+      "loss": 0.32628684,
+      "memory(GiB)": 33.01,
+      "step": 3790,
+      "train_speed(iter/s)": 0.187746
+    },
+    {
+      "acc": 0.923773,
+      "epoch": 1.257455268389662,
+      "grad_norm": 2.7790224059485245,
+      "learning_rate": 9.995470741562167e-06,
+      "loss": 0.34259708,
+      "memory(GiB)": 33.01,
+      "step": 3795,
+      "train_speed(iter/s)": 0.187753
+    },
+    {
+      "acc": 0.92042322,
+      "epoch": 1.259111994698476,
+      "grad_norm": 4.687488642901812,
+      "learning_rate": 9.995412271339736e-06,
+      "loss": 0.36016665,
+      "memory(GiB)": 33.01,
+      "step": 3800,
+      "train_speed(iter/s)": 0.187761
+    },
+    {
+      "acc": 0.92206593,
+      "epoch": 1.2607687210072895,
+      "grad_norm": 2.4026412571271343,
+      "learning_rate": 9.995353426298003e-06,
+      "loss": 0.39115138,
+      "memory(GiB)": 33.01,
+      "step": 3805,
+      "train_speed(iter/s)": 0.187768
+    },
+    {
+      "acc": 0.91857586,
+      "epoch": 1.2624254473161034,
+      "grad_norm": 4.513103775057416,
+      "learning_rate": 9.995294206441392e-06,
+      "loss": 0.36755357,
+      "memory(GiB)": 33.01,
+      "step": 3810,
+      "train_speed(iter/s)": 0.187775
+    },
+    {
+      "acc": 0.92359676,
+      "epoch": 1.2640821736249173,
+      "grad_norm": 3.262137887759917,
+      "learning_rate": 9.995234611774339e-06,
+      "loss": 0.32638867,
+      "memory(GiB)": 33.01,
+      "step": 3815,
+      "train_speed(iter/s)": 0.187781
+    },
+    {
+      "acc": 0.92083435,
+      "epoch": 1.265738899933731,
+      "grad_norm": 3.1376984773270946,
+      "learning_rate": 9.995174642301319e-06,
+      "loss": 0.39106462,
+      "memory(GiB)": 33.01,
+      "step": 3820,
+      "train_speed(iter/s)": 0.187788
+    },
+    {
+      "acc": 0.9268259,
+      "epoch": 1.2673956262425448,
+      "grad_norm": 3.6070613653945194,
+      "learning_rate": 9.995114298026833e-06,
+      "loss": 0.35192406,
+      "memory(GiB)": 33.01,
+      "step": 3825,
+      "train_speed(iter/s)": 0.187795
+    },
+    {
+      "acc": 0.92944641,
+      "epoch": 1.2690523525513586,
+      "grad_norm": 2.3387275683835504,
+      "learning_rate": 9.995053578955408e-06,
+      "loss": 0.31027389,
+      "memory(GiB)": 33.01,
+      "step": 3830,
+      "train_speed(iter/s)": 0.187802
+    },
+    {
+      "acc": 0.91999998,
+      "epoch": 1.2707090788601723,
+      "grad_norm": 3.382579511439133,
+      "learning_rate": 9.994992485091603e-06,
+      "loss": 0.35692229,
+      "memory(GiB)": 33.01,
+      "step": 3835,
+      "train_speed(iter/s)": 0.187808
+    },
+    {
+      "acc": 0.92861013,
+      "epoch": 1.2723658051689861,
+      "grad_norm": 3.6141616570107207,
+      "learning_rate": 9.994931016439998e-06,
+      "loss": 0.36012015,
+      "memory(GiB)": 33.01,
+      "step": 3840,
+      "train_speed(iter/s)": 0.187815
+    },
+    {
+      "acc": 0.92661362,
+      "epoch": 1.2740225314778,
+      "grad_norm": 2.7014468390409982,
+      "learning_rate": 9.99486917300521e-06,
+      "loss": 0.35332563,
+      "memory(GiB)": 33.01,
+      "step": 3845,
+      "train_speed(iter/s)": 0.187821
+    },
+    {
+      "acc": 0.93278389,
+      "epoch": 1.2756792577866136,
+      "grad_norm": 2.9190583238133536,
+      "learning_rate": 9.994806954791877e-06,
+      "loss": 0.32674026,
+      "memory(GiB)": 33.01,
+      "step": 3850,
+      "train_speed(iter/s)": 0.187828
+    },
+    {
+      "acc": 0.92834053,
+      "epoch": 1.2773359840954275,
+      "grad_norm": 3.162888980927606,
+      "learning_rate": 9.99474436180467e-06,
+      "loss": 0.35694146,
+      "memory(GiB)": 33.01,
+      "step": 3855,
+      "train_speed(iter/s)": 0.187834
+    },
+    {
+      "acc": 0.93106079,
+      "epoch": 1.2789927104042413,
+      "grad_norm": 2.180292217336007,
+      "learning_rate": 9.994681394048283e-06,
+      "loss": 0.31116631,
+      "memory(GiB)": 33.01,
+      "step": 3860,
+      "train_speed(iter/s)": 0.18784
+    },
+    {
+      "acc": 0.92896137,
+      "epoch": 1.280649436713055,
+      "grad_norm": 3.725631564421506,
+      "learning_rate": 9.994618051527444e-06,
+      "loss": 0.32231021,
+      "memory(GiB)": 33.01,
+      "step": 3865,
+      "train_speed(iter/s)": 0.187847
+    },
+    {
+      "acc": 0.928269,
+      "epoch": 1.2823061630218688,
+      "grad_norm": 3.8065518382089776,
+      "learning_rate": 9.994554334246902e-06,
+      "loss": 0.30769591,
+      "memory(GiB)": 33.01,
+      "step": 3870,
+      "train_speed(iter/s)": 0.187853
+    },
+    {
+      "acc": 0.93014402,
+      "epoch": 1.2839628893306827,
+      "grad_norm": 3.280700777914299,
+      "learning_rate": 9.994490242211445e-06,
+      "loss": 0.33068173,
+      "memory(GiB)": 33.01,
+      "step": 3875,
+      "train_speed(iter/s)": 0.18786
+    },
+    {
+      "acc": 0.93122234,
+      "epoch": 1.2856196156394963,
+      "grad_norm": 2.765686712261379,
+      "learning_rate": 9.994425775425877e-06,
+      "loss": 0.35716617,
+      "memory(GiB)": 33.01,
+      "step": 3880,
+      "train_speed(iter/s)": 0.187866
+    },
+    {
+      "acc": 0.93585377,
+      "epoch": 1.2872763419483102,
+      "grad_norm": 4.219149782391032,
+      "learning_rate": 9.994360933895037e-06,
+      "loss": 0.29821897,
+      "memory(GiB)": 33.01,
+      "step": 3885,
+      "train_speed(iter/s)": 0.187873
+    },
+    {
+      "acc": 0.94056263,
+      "epoch": 1.288933068257124,
+      "grad_norm": 3.7700987238795864,
+      "learning_rate": 9.994295717623792e-06,
+      "loss": 0.30137362,
+      "memory(GiB)": 33.01,
+      "step": 3890,
+      "train_speed(iter/s)": 0.187879
+    },
+    {
+      "acc": 0.92737751,
+      "epoch": 1.2905897945659377,
+      "grad_norm": 3.636527628612316,
+      "learning_rate": 9.994230126617035e-06,
+      "loss": 0.2921406,
+      "memory(GiB)": 33.01,
+      "step": 3895,
+      "train_speed(iter/s)": 0.187886
+    },
+    {
+      "acc": 0.93736925,
+      "epoch": 1.2922465208747516,
+      "grad_norm": 3.6644248366985575,
+      "learning_rate": 9.99416416087969e-06,
+      "loss": 0.29372125,
+      "memory(GiB)": 33.01,
+      "step": 3900,
+      "train_speed(iter/s)": 0.187892
+    },
+    {
+      "acc": 0.93469362,
+      "epoch": 1.2939032471835652,
+      "grad_norm": 3.5477819583404697,
+      "learning_rate": 9.994097820416702e-06,
+      "loss": 0.30780063,
+      "memory(GiB)": 33.01,
+      "step": 3905,
+      "train_speed(iter/s)": 0.187898
+    },
+    {
+      "acc": 0.92783298,
+      "epoch": 1.295559973492379,
+      "grad_norm": 3.141078965900213,
+      "learning_rate": 9.994031105233052e-06,
+      "loss": 0.31574149,
+      "memory(GiB)": 33.01,
+      "step": 3910,
+      "train_speed(iter/s)": 0.187905
+    },
+    {
+      "acc": 0.93738155,
+      "epoch": 1.297216699801193,
+      "grad_norm": 2.962093841370019,
+      "learning_rate": 9.993964015333749e-06,
+      "loss": 0.29025574,
+      "memory(GiB)": 33.01,
+      "step": 3915,
+      "train_speed(iter/s)": 0.187911
+    },
+    {
+      "acc": 0.92963943,
+      "epoch": 1.2988734261100066,
+      "grad_norm": 3.450074619131517,
+      "learning_rate": 9.993896550723824e-06,
+      "loss": 0.32397094,
+      "memory(GiB)": 33.01,
+      "step": 3920,
+      "train_speed(iter/s)": 0.187917
+    },
+    {
+      "acc": 0.92763071,
+      "epoch": 1.3005301524188204,
+      "grad_norm": 2.275170733239321,
+      "learning_rate": 9.993828711408341e-06,
+      "loss": 0.31115437,
+      "memory(GiB)": 33.01,
+      "step": 3925,
+      "train_speed(iter/s)": 0.187924
+    },
+    {
+      "acc": 0.93752851,
+      "epoch": 1.302186878727634,
+      "grad_norm": 2.4918963458658965,
+      "learning_rate": 9.993760497392388e-06,
+      "loss": 0.31371839,
+      "memory(GiB)": 33.01,
+      "step": 3930,
+      "train_speed(iter/s)": 0.18793
+    },
+    {
+      "acc": 0.9316061,
+      "epoch": 1.303843605036448,
+      "grad_norm": 3.053424190449544,
+      "learning_rate": 9.993691908681089e-06,
+      "loss": 0.35731153,
+      "memory(GiB)": 33.01,
+      "step": 3935,
+      "train_speed(iter/s)": 0.187936
+    },
+    {
+      "acc": 0.93001118,
+      "epoch": 1.3055003313452618,
+      "grad_norm": 2.388009315776182,
+      "learning_rate": 9.993622945279588e-06,
+      "loss": 0.2991662,
+      "memory(GiB)": 33.01,
+      "step": 3940,
+      "train_speed(iter/s)": 0.187942
+    },
+    {
+      "acc": 0.93689032,
+      "epoch": 1.3071570576540754,
+      "grad_norm": 2.625762219536255,
+      "learning_rate": 9.99355360719306e-06,
+      "loss": 0.30675516,
+      "memory(GiB)": 33.01,
+      "step": 3945,
+      "train_speed(iter/s)": 0.187949
+    },
+    {
+      "acc": 0.94865704,
+      "epoch": 1.3088137839628893,
+      "grad_norm": 2.9206118345721466,
+      "learning_rate": 9.993483894426706e-06,
+      "loss": 0.29642808,
+      "memory(GiB)": 33.01,
+      "step": 3950,
+      "train_speed(iter/s)": 0.187955
+    },
+    {
+      "acc": 0.92296667,
+      "epoch": 1.3104705102717031,
+      "grad_norm": 3.3467743607062808,
+      "learning_rate": 9.99341380698576e-06,
+      "loss": 0.33743124,
+      "memory(GiB)": 33.01,
+      "step": 3955,
+      "train_speed(iter/s)": 0.187962
+    },
+    {
+      "acc": 0.93078327,
+      "epoch": 1.3121272365805168,
+      "grad_norm": 2.1518445462143565,
+      "learning_rate": 9.993343344875482e-06,
+      "loss": 0.30151925,
+      "memory(GiB)": 33.01,
+      "step": 3960,
+      "train_speed(iter/s)": 0.187967
+    },
+    {
+      "acc": 0.93073835,
+      "epoch": 1.3137839628893306,
+      "grad_norm": 3.1234988518274767,
+      "learning_rate": 9.993272508101156e-06,
+      "loss": 0.29494073,
+      "memory(GiB)": 33.01,
+      "step": 3965,
+      "train_speed(iter/s)": 0.187973
+    },
+    {
+      "acc": 0.94187469,
+      "epoch": 1.3154406891981445,
+      "grad_norm": 2.884147113438224,
+      "learning_rate": 9.993201296668103e-06,
+      "loss": 0.30527532,
+      "memory(GiB)": 33.01,
+      "step": 3970,
+      "train_speed(iter/s)": 0.18798
+    },
+    {
+      "acc": 0.93317776,
+      "epoch": 1.3170974155069581,
+      "grad_norm": 2.596396758677392,
+      "learning_rate": 9.993129710581662e-06,
+      "loss": 0.28667531,
+      "memory(GiB)": 33.01,
+      "step": 3975,
+      "train_speed(iter/s)": 0.187986
+    },
+    {
+      "acc": 0.93754797,
+      "epoch": 1.318754141815772,
+      "grad_norm": 4.301438589905388,
+      "learning_rate": 9.993057749847207e-06,
+      "loss": 0.32379344,
+      "memory(GiB)": 33.01,
+      "step": 3980,
+      "train_speed(iter/s)": 0.187993
+    },
+    {
+      "acc": 0.92824039,
+      "epoch": 1.3204108681245859,
+      "grad_norm": 3.140141491914004,
+      "learning_rate": 9.992985414470136e-06,
+      "loss": 0.33532877,
+      "memory(GiB)": 33.01,
+      "step": 3985,
+      "train_speed(iter/s)": 0.187998
+    },
+    {
+      "acc": 0.93718243,
+      "epoch": 1.3220675944333995,
+      "grad_norm": 3.5344457077050038,
+      "learning_rate": 9.992912704455879e-06,
+      "loss": 0.36814497,
+      "memory(GiB)": 33.01,
+      "step": 3990,
+      "train_speed(iter/s)": 0.188004
+    },
+    {
+      "acc": 0.93041048,
+      "epoch": 1.3237243207422134,
+      "grad_norm": 2.7729199897251084,
+      "learning_rate": 9.992839619809893e-06,
+      "loss": 0.31639609,
+      "memory(GiB)": 33.01,
+      "step": 3995,
+      "train_speed(iter/s)": 0.188002
+    },
+    {
+      "acc": 0.93354034,
+      "epoch": 1.3253810470510272,
+      "grad_norm": 3.449036303659263,
+      "learning_rate": 9.992766160537659e-06,
+      "loss": 0.3051719,
+      "memory(GiB)": 33.01,
+      "step": 4000,
+      "train_speed(iter/s)": 0.188
+    },
+    {
+      "acc": 0.93086681,
+      "epoch": 1.3270377733598409,
+      "grad_norm": 3.0296237355416524,
+      "learning_rate": 9.992692326644693e-06,
+      "loss": 0.33369646,
+      "memory(GiB)": 33.01,
+      "step": 4005,
+      "train_speed(iter/s)": 0.188
+    },
+    {
+      "acc": 0.9316391,
+      "epoch": 1.3286944996686547,
+      "grad_norm": 2.5657778008931222,
+      "learning_rate": 9.992618118136533e-06,
+      "loss": 0.28221853,
+      "memory(GiB)": 33.01,
+      "step": 4010,
+      "train_speed(iter/s)": 0.188006
+    },
+    {
+      "acc": 0.92925653,
+      "epoch": 1.3303512259774686,
+      "grad_norm": 4.649549083512319,
+      "learning_rate": 9.99254353501875e-06,
+      "loss": 0.30690947,
+      "memory(GiB)": 33.01,
+      "step": 4015,
+      "train_speed(iter/s)": 0.188012
+    },
+    {
+      "acc": 0.94538603,
+      "epoch": 1.3320079522862822,
+      "grad_norm": 3.626225624410193,
+      "learning_rate": 9.992468577296935e-06,
+      "loss": 0.27149005,
+      "memory(GiB)": 33.01,
+      "step": 4020,
+      "train_speed(iter/s)": 0.188019
+    },
+    {
+      "acc": 0.94822292,
+      "epoch": 1.333664678595096,
+      "grad_norm": 3.855866149357719,
+      "learning_rate": 9.992393244976721e-06,
+      "loss": 0.26162114,
+      "memory(GiB)": 33.01,
+      "step": 4025,
+      "train_speed(iter/s)": 0.188025
+    },
+    {
+      "acc": 0.92598343,
+      "epoch": 1.33532140490391,
+      "grad_norm": 3.7938424779242004,
+      "learning_rate": 9.992317538063755e-06,
+      "loss": 0.34145732,
+      "memory(GiB)": 33.01,
+      "step": 4030,
+      "train_speed(iter/s)": 0.188031
+    },
+    {
+      "acc": 0.92380657,
+      "epoch": 1.3369781312127236,
+      "grad_norm": 4.716646594593983,
+      "learning_rate": 9.992241456563721e-06,
+      "loss": 0.38014688,
+      "memory(GiB)": 33.01,
+      "step": 4035,
+      "train_speed(iter/s)": 0.188037
+    },
+    {
+      "acc": 0.93459358,
+      "epoch": 1.3386348575215374,
+      "grad_norm": 4.612683598207418,
+      "learning_rate": 9.992165000482328e-06,
+      "loss": 0.2748986,
+      "memory(GiB)": 33.01,
+      "step": 4040,
+      "train_speed(iter/s)": 0.188042
+    },
+    {
+      "acc": 0.93223648,
+      "epoch": 1.3402915838303513,
+      "grad_norm": 2.592110016351656,
+      "learning_rate": 9.992088169825311e-06,
+      "loss": 0.32904916,
+      "memory(GiB)": 33.01,
+      "step": 4045,
+      "train_speed(iter/s)": 0.188049
+    },
+    {
+      "acc": 0.93688803,
+      "epoch": 1.341948310139165,
+      "grad_norm": 4.644288092052374,
+      "learning_rate": 9.992010964598436e-06,
+      "loss": 0.31696992,
+      "memory(GiB)": 33.01,
+      "step": 4050,
+      "train_speed(iter/s)": 0.188055
+    },
+    {
+      "acc": 0.93946381,
+      "epoch": 1.3436050364479788,
+      "grad_norm": 4.344444643704393,
+      "learning_rate": 9.9919333848075e-06,
+      "loss": 0.31000085,
+      "memory(GiB)": 33.01,
+      "step": 4055,
+      "train_speed(iter/s)": 0.188061
+    },
+    {
+      "acc": 0.93221197,
+      "epoch": 1.3452617627567927,
+      "grad_norm": 5.532366915026156,
+      "learning_rate": 9.991855430458322e-06,
+      "loss": 0.28737183,
+      "memory(GiB)": 33.01,
+      "step": 4060,
+      "train_speed(iter/s)": 0.188067
+    },
+    {
+      "acc": 0.94824142,
+      "epoch": 1.3469184890656063,
+      "grad_norm": 3.4428498074966236,
+      "learning_rate": 9.99177710155675e-06,
+      "loss": 0.28464704,
+      "memory(GiB)": 33.01,
+      "step": 4065,
+      "train_speed(iter/s)": 0.188072
+    },
+    {
+      "acc": 0.93260765,
+      "epoch": 1.3485752153744202,
+      "grad_norm": 4.935276047615211,
+      "learning_rate": 9.991698398108664e-06,
+      "loss": 0.29902873,
+      "memory(GiB)": 33.01,
+      "step": 4070,
+      "train_speed(iter/s)": 0.188078
+    },
+    {
+      "acc": 0.94271441,
+      "epoch": 1.350231941683234,
+      "grad_norm": 3.9581108800078924,
+      "learning_rate": 9.99161932011997e-06,
+      "loss": 0.27665021,
+      "memory(GiB)": 33.01,
+      "step": 4075,
+      "train_speed(iter/s)": 0.188084
+    },
+    {
+      "acc": 0.9331852,
+      "epoch": 1.3518886679920477,
+      "grad_norm": 3.4147382524913255,
+      "learning_rate": 9.9915398675966e-06,
+      "loss": 0.30838275,
+      "memory(GiB)": 33.01,
+      "step": 4080,
+      "train_speed(iter/s)": 0.18809
+    },
+    {
+      "acc": 0.93401146,
+      "epoch": 1.3535453943008615,
+      "grad_norm": 2.9195817564260866,
+      "learning_rate": 9.991460040544518e-06,
+      "loss": 0.29754572,
+      "memory(GiB)": 33.01,
+      "step": 4085,
+      "train_speed(iter/s)": 0.188096
+    },
+    {
+      "acc": 0.9369833,
+      "epoch": 1.3552021206096754,
+      "grad_norm": 3.8261472684027593,
+      "learning_rate": 9.991379838969713e-06,
+      "loss": 0.27704296,
+      "memory(GiB)": 33.01,
+      "step": 4090,
+      "train_speed(iter/s)": 0.188102
+    },
+    {
+      "acc": 0.93432493,
+      "epoch": 1.356858846918489,
+      "grad_norm": 3.0204807759996575,
+      "learning_rate": 9.991299262878208e-06,
+      "loss": 0.32643697,
+      "memory(GiB)": 33.01,
+      "step": 4095,
+      "train_speed(iter/s)": 0.188108
+    },
+    {
+      "acc": 0.93136787,
+      "epoch": 1.3585155732273029,
+      "grad_norm": 2.228315321411614,
+      "learning_rate": 9.991218312276043e-06,
+      "loss": 0.29976664,
+      "memory(GiB)": 33.01,
+      "step": 4100,
+      "train_speed(iter/s)": 0.188114
+    },
+    {
+      "acc": 0.93832722,
+      "epoch": 1.3601722995361167,
+      "grad_norm": 3.2780926082447217,
+      "learning_rate": 9.991136987169295e-06,
+      "loss": 0.31551151,
+      "memory(GiB)": 33.01,
+      "step": 4105,
+      "train_speed(iter/s)": 0.18812
+    },
+    {
+      "acc": 0.92250309,
+      "epoch": 1.3618290258449304,
+      "grad_norm": 2.0967913603524915,
+      "learning_rate": 9.991055287564067e-06,
+      "loss": 0.36131248,
+      "memory(GiB)": 33.01,
+      "step": 4110,
+      "train_speed(iter/s)": 0.188126
+    },
+    {
+      "acc": 0.9196558,
+      "epoch": 1.3634857521537442,
+      "grad_norm": 3.04957877993772,
+      "learning_rate": 9.99097321346649e-06,
+      "loss": 0.35065665,
+      "memory(GiB)": 33.01,
+      "step": 4115,
+      "train_speed(iter/s)": 0.188132
+    },
+    {
+      "acc": 0.93561573,
+      "epoch": 1.365142478462558,
+      "grad_norm": 3.0874861183406823,
+      "learning_rate": 9.990890764882722e-06,
+      "loss": 0.30160902,
+      "memory(GiB)": 33.01,
+      "step": 4120,
+      "train_speed(iter/s)": 0.188138
+    },
+    {
+      "acc": 0.93084764,
+      "epoch": 1.3667992047713717,
+      "grad_norm": 4.396726628372327,
+      "learning_rate": 9.990807941818951e-06,
+      "loss": 0.31668425,
+      "memory(GiB)": 33.01,
+      "step": 4125,
+      "train_speed(iter/s)": 0.188143
+    },
+    {
+      "acc": 0.94341707,
+      "epoch": 1.3684559310801856,
+      "grad_norm": 1.9439284972306012,
+      "learning_rate": 9.990724744281391e-06,
+      "loss": 0.2826818,
+      "memory(GiB)": 33.01,
+      "step": 4130,
+      "train_speed(iter/s)": 0.188149
+    },
+    {
+      "acc": 0.93765316,
+      "epoch": 1.3701126573889995,
+      "grad_norm": 1.8999731884417959,
+      "learning_rate": 9.990641172276286e-06,
+      "loss": 0.30255091,
+      "memory(GiB)": 33.01,
+      "step": 4135,
+      "train_speed(iter/s)": 0.188155
+    },
+    {
+      "acc": 0.9380127,
+      "epoch": 1.371769383697813,
+      "grad_norm": 2.873444423424632,
+      "learning_rate": 9.990557225809906e-06,
+      "loss": 0.28472929,
+      "memory(GiB)": 33.01,
+      "step": 4140,
+      "train_speed(iter/s)": 0.188161
+    },
+    {
+      "acc": 0.94341812,
+      "epoch": 1.373426110006627,
+      "grad_norm": 2.7763226622387758,
+      "learning_rate": 9.990472904888552e-06,
+      "loss": 0.27308426,
+      "memory(GiB)": 33.01,
+      "step": 4145,
+      "train_speed(iter/s)": 0.188167
+    },
+    {
+      "acc": 0.94635715,
+      "epoch": 1.3750828363154408,
+      "grad_norm": 2.5908085822084534,
+      "learning_rate": 9.990388209518552e-06,
+      "loss": 0.26468968,
+      "memory(GiB)": 33.01,
+      "step": 4150,
+      "train_speed(iter/s)": 0.188172
+    },
+    {
+      "acc": 0.93955574,
+      "epoch": 1.3767395626242545,
+      "grad_norm": 4.805493021778142,
+      "learning_rate": 9.990303139706258e-06,
+      "loss": 0.29098492,
+      "memory(GiB)": 33.01,
+      "step": 4155,
+      "train_speed(iter/s)": 0.188178
+    },
+    {
+      "acc": 0.92813005,
+      "epoch": 1.3783962889330683,
+      "grad_norm": 3.9214892603691798,
+      "learning_rate": 9.990217695458058e-06,
+      "loss": 0.3426239,
+      "memory(GiB)": 33.01,
+      "step": 4160,
+      "train_speed(iter/s)": 0.188184
+    },
+    {
+      "acc": 0.91839867,
+      "epoch": 1.3800530152418822,
+      "grad_norm": 4.045315242690874,
+      "learning_rate": 9.99013187678036e-06,
+      "loss": 0.3696887,
+      "memory(GiB)": 33.01,
+      "step": 4165,
+      "train_speed(iter/s)": 0.188189
+    },
+    {
+      "acc": 0.9324028,
+      "epoch": 1.3817097415506958,
+      "grad_norm": 4.321959254473603,
+      "learning_rate": 9.990045683679605e-06,
+      "loss": 0.31572053,
+      "memory(GiB)": 33.01,
+      "step": 4170,
+      "train_speed(iter/s)": 0.188195
+    },
+    {
+      "acc": 0.92923403,
+      "epoch": 1.3833664678595097,
+      "grad_norm": 2.3322544989628042,
+      "learning_rate": 9.989959116162263e-06,
+      "loss": 0.33993003,
+      "memory(GiB)": 33.01,
+      "step": 4175,
+      "train_speed(iter/s)": 0.188201
+    },
+    {
+      "acc": 0.9367238,
+      "epoch": 1.3850231941683233,
+      "grad_norm": 2.451834543607949,
+      "learning_rate": 9.989872174234829e-06,
+      "loss": 0.2900305,
+      "memory(GiB)": 33.01,
+      "step": 4180,
+      "train_speed(iter/s)": 0.188206
+    },
+    {
+      "acc": 0.93373671,
+      "epoch": 1.3866799204771372,
+      "grad_norm": 2.745893558754339,
+      "learning_rate": 9.989784857903826e-06,
+      "loss": 0.30362544,
+      "memory(GiB)": 33.01,
+      "step": 4185,
+      "train_speed(iter/s)": 0.188211
+    },
+    {
+      "acc": 0.94090681,
+      "epoch": 1.388336646785951,
+      "grad_norm": 2.333242180787866,
+      "learning_rate": 9.989697167175807e-06,
+      "loss": 0.29308481,
+      "memory(GiB)": 33.01,
+      "step": 4190,
+      "train_speed(iter/s)": 0.188217
+    },
+    {
+      "acc": 0.93064957,
+      "epoch": 1.3899933730947647,
+      "grad_norm": 2.676326656969343,
+      "learning_rate": 9.989609102057351e-06,
+      "loss": 0.3038188,
+      "memory(GiB)": 33.01,
+      "step": 4195,
+      "train_speed(iter/s)": 0.188223
+    },
+    {
+      "acc": 0.93896952,
+      "epoch": 1.3916500994035785,
+      "grad_norm": 2.926888898906929,
+      "learning_rate": 9.98952066255507e-06,
+      "loss": 0.31832986,
+      "memory(GiB)": 33.01,
+      "step": 4200,
+      "train_speed(iter/s)": 0.188228
+    },
+    {
+      "acc": 0.93322506,
+      "epoch": 1.3933068257123924,
+      "grad_norm": 1.9729320822722771,
+      "learning_rate": 9.989431848675594e-06,
+      "loss": 0.31466131,
+      "memory(GiB)": 33.01,
+      "step": 4205,
+      "train_speed(iter/s)": 0.188234
+    },
+    {
+      "acc": 0.9312398,
+      "epoch": 1.394963552021206,
+      "grad_norm": 2.4255073560346796,
+      "learning_rate": 9.989342660425596e-06,
+      "loss": 0.31000106,
+      "memory(GiB)": 33.01,
+      "step": 4210,
+      "train_speed(iter/s)": 0.18824
+    },
+    {
+      "acc": 0.93309641,
+      "epoch": 1.39662027833002,
+      "grad_norm": 3.1566325185723803,
+      "learning_rate": 9.989253097811762e-06,
+      "loss": 0.2976181,
+      "memory(GiB)": 33.01,
+      "step": 4215,
+      "train_speed(iter/s)": 0.188245
+    },
+    {
+      "acc": 0.93946476,
+      "epoch": 1.3982770046388335,
+      "grad_norm": 2.476569856253078,
+      "learning_rate": 9.989163160840818e-06,
+      "loss": 0.28907919,
+      "memory(GiB)": 33.01,
+      "step": 4220,
+      "train_speed(iter/s)": 0.188248
+    },
+    {
+      "acc": 0.94837618,
+      "epoch": 1.3999337309476474,
+      "grad_norm": 2.911679720820383,
+      "learning_rate": 9.989072849519508e-06,
+      "loss": 0.27384791,
+      "memory(GiB)": 33.01,
+      "step": 4225,
+      "train_speed(iter/s)": 0.188248
+    },
+    {
+      "acc": 0.93348541,
+      "epoch": 1.4015904572564613,
+      "grad_norm": 3.441434835167615,
+      "learning_rate": 9.988982163854613e-06,
+      "loss": 0.28337784,
+      "memory(GiB)": 33.01,
+      "step": 4230,
+      "train_speed(iter/s)": 0.188243
+    },
+    {
+      "acc": 0.93422871,
+      "epoch": 1.403247183565275,
+      "grad_norm": 3.1333176002891996,
+      "learning_rate": 9.988891103852935e-06,
+      "loss": 0.28754125,
+      "memory(GiB)": 33.01,
+      "step": 4235,
+      "train_speed(iter/s)": 0.188248
+    },
+    {
+      "acc": 0.94224987,
+      "epoch": 1.4049039098740888,
+      "grad_norm": 3.86832950886018,
+      "learning_rate": 9.988799669521308e-06,
+      "loss": 0.28542409,
+      "memory(GiB)": 33.01,
+      "step": 4240,
+      "train_speed(iter/s)": 0.188254
+    },
+    {
+      "acc": 0.94306545,
+      "epoch": 1.4065606361829026,
+      "grad_norm": 2.243059423935416,
+      "learning_rate": 9.988707860866595e-06,
+      "loss": 0.26384854,
+      "memory(GiB)": 33.01,
+      "step": 4245,
+      "train_speed(iter/s)": 0.18826
+    },
+    {
+      "acc": 0.94372959,
+      "epoch": 1.4082173624917163,
+      "grad_norm": 2.4386320268099504,
+      "learning_rate": 9.988615677895683e-06,
+      "loss": 0.27046497,
+      "memory(GiB)": 33.01,
+      "step": 4250,
+      "train_speed(iter/s)": 0.188265
+    },
+    {
+      "acc": 0.94478703,
+      "epoch": 1.4098740888005301,
+      "grad_norm": 2.866006036158438,
+      "learning_rate": 9.98852312061549e-06,
+      "loss": 0.26427975,
+      "memory(GiB)": 33.01,
+      "step": 4255,
+      "train_speed(iter/s)": 0.18827
+    },
+    {
+      "acc": 0.94633064,
+      "epoch": 1.411530815109344,
+      "grad_norm": 2.996758179153441,
+      "learning_rate": 9.988430189032962e-06,
+      "loss": 0.26404467,
+      "memory(GiB)": 33.01,
+      "step": 4260,
+      "train_speed(iter/s)": 0.188276
+    },
+    {
+      "acc": 0.94439745,
+      "epoch": 1.4131875414181576,
+      "grad_norm": 3.431455227332249,
+      "learning_rate": 9.988336883155075e-06,
+      "loss": 0.27302637,
+      "memory(GiB)": 33.01,
+      "step": 4265,
+      "train_speed(iter/s)": 0.188281
+    },
+    {
+      "acc": 0.93996572,
+      "epoch": 1.4148442677269715,
+      "grad_norm": 3.5621623685509443,
+      "learning_rate": 9.988243202988825e-06,
+      "loss": 0.28133731,
+      "memory(GiB)": 33.01,
+      "step": 4270,
+      "train_speed(iter/s)": 0.188287
+    },
+    {
+      "acc": 0.93899097,
+      "epoch": 1.4165009940357853,
+      "grad_norm": 2.576958716798402,
+      "learning_rate": 9.988149148541247e-06,
+      "loss": 0.30188389,
+      "memory(GiB)": 33.01,
+      "step": 4275,
+      "train_speed(iter/s)": 0.188293
+    },
+    {
+      "acc": 0.93714771,
+      "epoch": 1.418157720344599,
+      "grad_norm": 3.6469486042808006,
+      "learning_rate": 9.988054719819396e-06,
+      "loss": 0.28270049,
+      "memory(GiB)": 33.01,
+      "step": 4280,
+      "train_speed(iter/s)": 0.188299
+    },
+    {
+      "acc": 0.93806877,
+      "epoch": 1.4198144466534128,
+      "grad_norm": 4.028564160520423,
+      "learning_rate": 9.987959916830359e-06,
+      "loss": 0.34861684,
+      "memory(GiB)": 33.01,
+      "step": 4285,
+      "train_speed(iter/s)": 0.188304
+    },
+    {
+      "acc": 0.93435154,
+      "epoch": 1.4214711729622267,
+      "grad_norm": 2.7449103891634503,
+      "learning_rate": 9.98786473958125e-06,
+      "loss": 0.2720546,
+      "memory(GiB)": 33.01,
+      "step": 4290,
+      "train_speed(iter/s)": 0.188309
+    },
+    {
+      "acc": 0.94607534,
+      "epoch": 1.4231278992710403,
+      "grad_norm": 3.7844318142821773,
+      "learning_rate": 9.987769188079212e-06,
+      "loss": 0.28037467,
+      "memory(GiB)": 33.01,
+      "step": 4295,
+      "train_speed(iter/s)": 0.188314
+    },
+    {
+      "acc": 0.93993568,
+      "epoch": 1.4247846255798542,
+      "grad_norm": 5.149616419195095,
+      "learning_rate": 9.987673262331412e-06,
+      "loss": 0.28110685,
+      "memory(GiB)": 33.01,
+      "step": 4300,
+      "train_speed(iter/s)": 0.18832
+    },
+    {
+      "acc": 0.93993053,
+      "epoch": 1.426441351888668,
+      "grad_norm": 4.1529337866853915,
+      "learning_rate": 9.987576962345053e-06,
+      "loss": 0.28562617,
+      "memory(GiB)": 33.01,
+      "step": 4305,
+      "train_speed(iter/s)": 0.188325
+    },
+    {
+      "acc": 0.95176363,
+      "epoch": 1.4280980781974817,
+      "grad_norm": 3.5657578833957273,
+      "learning_rate": 9.987480288127357e-06,
+      "loss": 0.23848631,
+      "memory(GiB)": 33.01,
+      "step": 4310,
+      "train_speed(iter/s)": 0.188331
+    },
+    {
+      "acc": 0.94450035,
+      "epoch": 1.4297548045062956,
+      "grad_norm": 3.189509654070219,
+      "learning_rate": 9.987383239685581e-06,
+      "loss": 0.26595705,
+      "memory(GiB)": 33.01,
+      "step": 4315,
+      "train_speed(iter/s)": 0.188337
+    },
+    {
+      "acc": 0.94286051,
+      "epoch": 1.4314115308151094,
+      "grad_norm": 2.614431557513013,
+      "learning_rate": 9.987285817027007e-06,
+      "loss": 0.2524662,
+      "memory(GiB)": 33.01,
+      "step": 4320,
+      "train_speed(iter/s)": 0.188342
+    },
+    {
+      "acc": 0.94530735,
+      "epoch": 1.433068257123923,
+      "grad_norm": 4.132661227795143,
+      "learning_rate": 9.987188020158944e-06,
+      "loss": 0.24585824,
+      "memory(GiB)": 33.01,
+      "step": 4325,
+      "train_speed(iter/s)": 0.188347
+    },
+    {
+      "acc": 0.93835888,
+      "epoch": 1.434724983432737,
+      "grad_norm": 4.110450519736193,
+      "learning_rate": 9.987089849088735e-06,
+      "loss": 0.28327467,
+      "memory(GiB)": 33.01,
+      "step": 4330,
+      "train_speed(iter/s)": 0.188353
+    },
+    {
+      "acc": 0.94487133,
+      "epoch": 1.4363817097415508,
+      "grad_norm": 3.046038884222029,
+      "learning_rate": 9.986991303823742e-06,
+      "loss": 0.25465138,
+      "memory(GiB)": 33.01,
+      "step": 4335,
+      "train_speed(iter/s)": 0.188358
+    },
+    {
+      "acc": 0.95449142,
+      "epoch": 1.4380384360503644,
+      "grad_norm": 3.1566223779421487,
+      "learning_rate": 9.986892384371363e-06,
+      "loss": 0.22105565,
+      "memory(GiB)": 33.01,
+      "step": 4340,
+      "train_speed(iter/s)": 0.188363
+    },
+    {
+      "acc": 0.94515934,
+      "epoch": 1.4396951623591783,
+      "grad_norm": 3.1217568742644777,
+      "learning_rate": 9.986793090739018e-06,
+      "loss": 0.25169601,
+      "memory(GiB)": 33.01,
+      "step": 4345,
+      "train_speed(iter/s)": 0.188369
+    },
+    {
+      "acc": 0.94277983,
+      "epoch": 1.4413518886679921,
+      "grad_norm": 3.596221110177539,
+      "learning_rate": 9.98669342293416e-06,
+      "loss": 0.2747798,
+      "memory(GiB)": 33.01,
+      "step": 4350,
+      "train_speed(iter/s)": 0.188374
+    },
+    {
+      "acc": 0.9469965,
+      "epoch": 1.4430086149768058,
+      "grad_norm": 2.9384089218959044,
+      "learning_rate": 9.986593380964269e-06,
+      "loss": 0.25603518,
+      "memory(GiB)": 33.01,
+      "step": 4355,
+      "train_speed(iter/s)": 0.188379
+    },
+    {
+      "acc": 0.94336195,
+      "epoch": 1.4446653412856196,
+      "grad_norm": 3.4666269364313886,
+      "learning_rate": 9.986492964836851e-06,
+      "loss": 0.2590342,
+      "memory(GiB)": 33.01,
+      "step": 4360,
+      "train_speed(iter/s)": 0.188384
+    },
+    {
+      "acc": 0.94749794,
+      "epoch": 1.4463220675944335,
+      "grad_norm": 2.4064228589733814,
+      "learning_rate": 9.986392174559441e-06,
+      "loss": 0.24870133,
+      "memory(GiB)": 33.01,
+      "step": 4365,
+      "train_speed(iter/s)": 0.188389
+    },
+    {
+      "acc": 0.93699341,
+      "epoch": 1.4479787939032471,
+      "grad_norm": 2.677875073448861,
+      "learning_rate": 9.986291010139601e-06,
+      "loss": 0.29232826,
+      "memory(GiB)": 33.01,
+      "step": 4370,
+      "train_speed(iter/s)": 0.188395
+    },
+    {
+      "acc": 0.93877039,
+      "epoch": 1.449635520212061,
+      "grad_norm": 2.8612888959465925,
+      "learning_rate": 9.986189471584927e-06,
+      "loss": 0.27531915,
+      "memory(GiB)": 33.01,
+      "step": 4375,
+      "train_speed(iter/s)": 0.1884
+    },
+    {
+      "acc": 0.9468504,
+      "epoch": 1.4512922465208749,
+      "grad_norm": 2.9523552229828667,
+      "learning_rate": 9.986087558903034e-06,
+      "loss": 0.24648032,
+      "memory(GiB)": 33.01,
+      "step": 4380,
+      "train_speed(iter/s)": 0.188405
+    },
+    {
+      "acc": 0.93813934,
+      "epoch": 1.4529489728296885,
+      "grad_norm": 3.0496526021997323,
+      "learning_rate": 9.985985272101572e-06,
+      "loss": 0.28014739,
+      "memory(GiB)": 33.01,
+      "step": 4385,
+      "train_speed(iter/s)": 0.18841
+    },
+    {
+      "acc": 0.94850588,
+      "epoch": 1.4546056991385024,
+      "grad_norm": 2.9243739816215273,
+      "learning_rate": 9.985882611188215e-06,
+      "loss": 0.2484787,
+      "memory(GiB)": 33.01,
+      "step": 4390,
+      "train_speed(iter/s)": 0.188416
+    },
+    {
+      "acc": 0.93145809,
+      "epoch": 1.4562624254473162,
+      "grad_norm": 2.603042420356731,
+      "learning_rate": 9.985779576170667e-06,
+      "loss": 0.30901597,
+      "memory(GiB)": 33.01,
+      "step": 4395,
+      "train_speed(iter/s)": 0.18842
+    },
+    {
+      "acc": 0.94351511,
+      "epoch": 1.4579191517561298,
+      "grad_norm": 2.289743784223293,
+      "learning_rate": 9.98567616705666e-06,
+      "loss": 0.26875901,
+      "memory(GiB)": 33.01,
+      "step": 4400,
+      "train_speed(iter/s)": 0.188425
+    },
+    {
+      "acc": 0.94040852,
+      "epoch": 1.4595758780649437,
+      "grad_norm": 4.596671413802325,
+      "learning_rate": 9.985572383853952e-06,
+      "loss": 0.2778924,
+      "memory(GiB)": 33.01,
+      "step": 4405,
+      "train_speed(iter/s)": 0.18843
+    },
+    {
+      "acc": 0.93312197,
+      "epoch": 1.4612326043737576,
+      "grad_norm": 3.280940498392058,
+      "learning_rate": 9.985468226570338e-06,
+      "loss": 0.30632141,
+      "memory(GiB)": 33.01,
+      "step": 4410,
+      "train_speed(iter/s)": 0.188435
+    },
+    {
+      "acc": 0.94229469,
+      "epoch": 1.4628893306825712,
+      "grad_norm": 2.5384946557496315,
+      "learning_rate": 9.985363695213625e-06,
+      "loss": 0.28237987,
+      "memory(GiB)": 33.01,
+      "step": 4415,
+      "train_speed(iter/s)": 0.18844
+    },
+    {
+      "acc": 0.9361887,
+      "epoch": 1.464546056991385,
+      "grad_norm": 2.810763212581721,
+      "learning_rate": 9.985258789791661e-06,
+      "loss": 0.29882345,
+      "memory(GiB)": 33.01,
+      "step": 4420,
+      "train_speed(iter/s)": 0.188445
+    },
+    {
+      "acc": 0.94827404,
+      "epoch": 1.466202783300199,
+      "grad_norm": 2.606906569992161,
+      "learning_rate": 9.98515351031232e-06,
+      "loss": 0.25767074,
+      "memory(GiB)": 33.01,
+      "step": 4425,
+      "train_speed(iter/s)": 0.18845
+    },
+    {
+      "acc": 0.94562292,
+      "epoch": 1.4678595096090126,
+      "grad_norm": 4.317781725106929,
+      "learning_rate": 9.9850478567835e-06,
+      "loss": 0.25841742,
+      "memory(GiB)": 33.01,
+      "step": 4430,
+      "train_speed(iter/s)": 0.188456
+    },
+    {
+      "acc": 0.94369373,
+      "epoch": 1.4695162359178264,
+      "grad_norm": 3.1070989538697598,
+      "learning_rate": 9.984941829213128e-06,
+      "loss": 0.26027231,
+      "memory(GiB)": 33.01,
+      "step": 4435,
+      "train_speed(iter/s)": 0.18846
+    },
+    {
+      "acc": 0.93526039,
+      "epoch": 1.4711729622266403,
+      "grad_norm": 3.0051770344683617,
+      "learning_rate": 9.984835427609161e-06,
+      "loss": 0.32055621,
+      "memory(GiB)": 33.01,
+      "step": 4440,
+      "train_speed(iter/s)": 0.188466
+    },
+    {
+      "acc": 0.94421978,
+      "epoch": 1.472829688535454,
+      "grad_norm": 3.3700092652183886,
+      "learning_rate": 9.984728651979587e-06,
+      "loss": 0.26462595,
+      "memory(GiB)": 33.01,
+      "step": 4445,
+      "train_speed(iter/s)": 0.188471
+    },
+    {
+      "acc": 0.94748573,
+      "epoch": 1.4744864148442678,
+      "grad_norm": 2.433854802806776,
+      "learning_rate": 9.984621502332415e-06,
+      "loss": 0.22584906,
+      "memory(GiB)": 33.01,
+      "step": 4450,
+      "train_speed(iter/s)": 0.188468
+    },
+    {
+      "acc": 0.93821697,
+      "epoch": 1.4761431411530814,
+      "grad_norm": 3.3424554831160886,
+      "learning_rate": 9.984513978675685e-06,
+      "loss": 0.2816637,
+      "memory(GiB)": 33.01,
+      "step": 4455,
+      "train_speed(iter/s)": 0.188473
+    },
+    {
+      "acc": 0.94929428,
+      "epoch": 1.4777998674618953,
+      "grad_norm": 2.2572600206045546,
+      "learning_rate": 9.984406081017468e-06,
+      "loss": 0.23356147,
+      "memory(GiB)": 33.01,
+      "step": 4460,
+      "train_speed(iter/s)": 0.188469
+    },
+    {
+      "acc": 0.94652576,
+      "epoch": 1.4794565937707091,
+      "grad_norm": 3.575363670522039,
+      "learning_rate": 9.984297809365858e-06,
+      "loss": 0.24346108,
+      "memory(GiB)": 33.01,
+      "step": 4465,
+      "train_speed(iter/s)": 0.188474
+    },
+    {
+      "acc": 0.9413991,
+      "epoch": 1.4811133200795228,
+      "grad_norm": 3.0839640458880244,
+      "learning_rate": 9.984189163728982e-06,
+      "loss": 0.27495437,
+      "memory(GiB)": 33.01,
+      "step": 4470,
+      "train_speed(iter/s)": 0.188479
+    },
+    {
+      "acc": 0.93963642,
+      "epoch": 1.4827700463883366,
+      "grad_norm": 3.111673666441867,
+      "learning_rate": 9.984080144114992e-06,
+      "loss": 0.25605507,
+      "memory(GiB)": 33.01,
+      "step": 4475,
+      "train_speed(iter/s)": 0.188484
+    },
+    {
+      "acc": 0.93898993,
+      "epoch": 1.4844267726971505,
+      "grad_norm": 3.0321815813549273,
+      "learning_rate": 9.98397075053207e-06,
+      "loss": 0.30646503,
+      "memory(GiB)": 33.01,
+      "step": 4480,
+      "train_speed(iter/s)": 0.188489
+    },
+    {
+      "acc": 0.95032883,
+      "epoch": 1.4860834990059641,
+      "grad_norm": 2.156852537631904,
+      "learning_rate": 9.983860982988422e-06,
+      "loss": 0.22426569,
+      "memory(GiB)": 33.01,
+      "step": 4485,
+      "train_speed(iter/s)": 0.188494
+    },
+    {
+      "acc": 0.94717827,
+      "epoch": 1.487740225314778,
+      "grad_norm": 2.5987339576925934,
+      "learning_rate": 9.983750841492289e-06,
+      "loss": 0.25280399,
+      "memory(GiB)": 33.01,
+      "step": 4490,
+      "train_speed(iter/s)": 0.188499
+    },
+    {
+      "acc": 0.93444853,
+      "epoch": 1.4893969516235916,
+      "grad_norm": 3.8845195701504243,
+      "learning_rate": 9.983640326051933e-06,
+      "loss": 0.31441751,
+      "memory(GiB)": 33.01,
+      "step": 4495,
+      "train_speed(iter/s)": 0.188504
+    },
+    {
+      "acc": 0.94258862,
+      "epoch": 1.4910536779324055,
+      "grad_norm": 3.844360759369666,
+      "learning_rate": 9.983529436675647e-06,
+      "loss": 0.26622539,
+      "memory(GiB)": 33.01,
+      "step": 4500,
+      "train_speed(iter/s)": 0.188509
+    },
+    {
+      "acc": 0.94434681,
+      "epoch": 1.4927104042412194,
+      "grad_norm": 2.918815300404156,
+      "learning_rate": 9.983418173371753e-06,
+      "loss": 0.31492476,
+      "memory(GiB)": 33.01,
+      "step": 4505,
+      "train_speed(iter/s)": 0.188514
+    },
+    {
+      "acc": 0.96208019,
+      "epoch": 1.494367130550033,
+      "grad_norm": 2.5202193462964795,
+      "learning_rate": 9.9833065361486e-06,
+      "loss": 0.21308544,
+      "memory(GiB)": 33.01,
+      "step": 4510,
+      "train_speed(iter/s)": 0.188519
+    },
+    {
+      "acc": 0.94232941,
+      "epoch": 1.4960238568588469,
+      "grad_norm": 3.166034806246495,
+      "learning_rate": 9.983194525014566e-06,
+      "loss": 0.2875102,
+      "memory(GiB)": 33.01,
+      "step": 4515,
+      "train_speed(iter/s)": 0.188523
+    },
+    {
+      "acc": 0.92990093,
+      "epoch": 1.4976805831676607,
+      "grad_norm": 5.907539194923817,
+      "learning_rate": 9.983082139978058e-06,
+      "loss": 0.32663989,
+      "memory(GiB)": 33.01,
+      "step": 4520,
+      "train_speed(iter/s)": 0.188528
+    },
+    {
+      "acc": 0.94061928,
+      "epoch": 1.4993373094764744,
+      "grad_norm": 3.8378386950851673,
+      "learning_rate": 9.982969381047506e-06,
+      "loss": 0.27804546,
+      "memory(GiB)": 33.01,
+      "step": 4525,
+      "train_speed(iter/s)": 0.188533
+    },
+    {
+      "acc": 0.9455164,
+      "epoch": 1.5009940357852882,
+      "grad_norm": 2.4790047190754887,
+      "learning_rate": 9.982856248231372e-06,
+      "loss": 0.26023202,
+      "memory(GiB)": 33.01,
+      "step": 4530,
+      "train_speed(iter/s)": 0.188538
+    },
+    {
+      "acc": 0.94345798,
+      "epoch": 1.502650762094102,
+      "grad_norm": 3.0573407951814806,
+      "learning_rate": 9.982742741538146e-06,
+      "loss": 0.2570219,
+      "memory(GiB)": 33.01,
+      "step": 4535,
+      "train_speed(iter/s)": 0.188543
+    },
+    {
+      "acc": 0.94834957,
+      "epoch": 1.5043074884029157,
+      "grad_norm": 4.104025784557431,
+      "learning_rate": 9.982628860976346e-06,
+      "loss": 0.25655692,
+      "memory(GiB)": 33.01,
+      "step": 4540,
+      "train_speed(iter/s)": 0.188548
+    },
+    {
+      "acc": 0.95639915,
+      "epoch": 1.5059642147117296,
+      "grad_norm": 3.93063717312741,
+      "learning_rate": 9.982514606554518e-06,
+      "loss": 0.22867155,
+      "memory(GiB)": 33.01,
+      "step": 4545,
+      "train_speed(iter/s)": 0.188553
+    },
+    {
+      "acc": 0.95113354,
+      "epoch": 1.5076209410205434,
+      "grad_norm": 20.94903292322898,
+      "learning_rate": 9.982399978281236e-06,
+      "loss": 0.27273102,
+      "memory(GiB)": 33.01,
+      "step": 4550,
+      "train_speed(iter/s)": 0.188558
+    },
+    {
+      "acc": 0.95819855,
+      "epoch": 1.509277667329357,
+      "grad_norm": 3.4387520189083847,
+      "learning_rate": 9.9822849761651e-06,
+      "loss": 0.23380876,
+      "memory(GiB)": 33.01,
+      "step": 4555,
+      "train_speed(iter/s)": 0.188563
+    },
+    {
+      "acc": 0.95431976,
+      "epoch": 1.510934393638171,
+      "grad_norm": 3.562322457355702,
+      "learning_rate": 9.98216960021474e-06,
+      "loss": 0.20716219,
+      "memory(GiB)": 33.01,
+      "step": 4560,
+      "train_speed(iter/s)": 0.188568
+    },
+    {
+      "acc": 0.95186443,
+      "epoch": 1.5125911199469848,
+      "grad_norm": 3.680971429469887,
+      "learning_rate": 9.982053850438816e-06,
+      "loss": 0.22953637,
+      "memory(GiB)": 33.01,
+      "step": 4565,
+      "train_speed(iter/s)": 0.188573
+    },
+    {
+      "acc": 0.95531559,
+      "epoch": 1.5142478462557984,
+      "grad_norm": 4.100797831635736,
+      "learning_rate": 9.981937726846012e-06,
+      "loss": 0.22091107,
+      "memory(GiB)": 33.01,
+      "step": 4570,
+      "train_speed(iter/s)": 0.188578
+    },
+    {
+      "acc": 0.94186716,
+      "epoch": 1.5159045725646123,
+      "grad_norm": 3.3223197471269157,
+      "learning_rate": 9.981821229445041e-06,
+      "loss": 0.26601729,
+      "memory(GiB)": 33.01,
+      "step": 4575,
+      "train_speed(iter/s)": 0.188583
+    },
+    {
+      "acc": 0.93858147,
+      "epoch": 1.5175612988734262,
+      "grad_norm": 3.3266660382792055,
+      "learning_rate": 9.981704358244647e-06,
+      "loss": 0.28532922,
+      "memory(GiB)": 33.01,
+      "step": 4580,
+      "train_speed(iter/s)": 0.188588
+    },
+    {
+      "acc": 0.94062061,
+      "epoch": 1.5192180251822398,
+      "grad_norm": 2.8894966939472817,
+      "learning_rate": 9.9815871132536e-06,
+      "loss": 0.29712996,
+      "memory(GiB)": 33.01,
+      "step": 4585,
+      "train_speed(iter/s)": 0.188593
+    },
+    {
+      "acc": 0.94864483,
+      "epoch": 1.5208747514910537,
+      "grad_norm": 2.538374198617406,
+      "learning_rate": 9.981469494480699e-06,
+      "loss": 0.23360934,
+      "memory(GiB)": 33.01,
+      "step": 4590,
+      "train_speed(iter/s)": 0.188597
+    },
+    {
+      "acc": 0.93387566,
+      "epoch": 1.5225314777998675,
+      "grad_norm": 3.4944236423447506,
+      "learning_rate": 9.981351501934764e-06,
+      "loss": 0.31100078,
+      "memory(GiB)": 33.01,
+      "step": 4595,
+      "train_speed(iter/s)": 0.188602
+    },
+    {
+      "acc": 0.95249996,
+      "epoch": 1.5241882041086812,
+      "grad_norm": 3.2114280847773164,
+      "learning_rate": 9.981233135624658e-06,
+      "loss": 0.2311841,
+      "memory(GiB)": 33.01,
+      "step": 4600,
+      "train_speed(iter/s)": 0.188607
+    },
+    {
+      "acc": 0.94958639,
+      "epoch": 1.525844930417495,
+      "grad_norm": 2.3210712199238377,
+      "learning_rate": 9.981114395559257e-06,
+      "loss": 0.24990735,
+      "memory(GiB)": 33.01,
+      "step": 4605,
+      "train_speed(iter/s)": 0.188611
+    },
+    {
+      "acc": 0.95304022,
+      "epoch": 1.5275016567263089,
+      "grad_norm": 2.128002424742902,
+      "learning_rate": 9.980995281747473e-06,
+      "loss": 0.223489,
+      "memory(GiB)": 33.01,
+      "step": 4610,
+      "train_speed(iter/s)": 0.188617
+    },
+    {
+      "acc": 0.94253674,
+      "epoch": 1.5291583830351225,
+      "grad_norm": 2.872106254199602,
+      "learning_rate": 9.980875794198245e-06,
+      "loss": 0.28155315,
+      "memory(GiB)": 33.01,
+      "step": 4615,
+      "train_speed(iter/s)": 0.188621
+    },
+    {
+      "acc": 0.94812603,
+      "epoch": 1.5308151093439364,
+      "grad_norm": 3.301666913276626,
+      "learning_rate": 9.98075593292054e-06,
+      "loss": 0.24484925,
+      "memory(GiB)": 33.01,
+      "step": 4620,
+      "train_speed(iter/s)": 0.188626
+    },
+    {
+      "acc": 0.94433002,
+      "epoch": 1.5324718356527502,
+      "grad_norm": 3.023955619792361,
+      "learning_rate": 9.98063569792335e-06,
+      "loss": 0.27217543,
+      "memory(GiB)": 33.01,
+      "step": 4625,
+      "train_speed(iter/s)": 0.188631
+    },
+    {
+      "acc": 0.94974155,
+      "epoch": 1.5341285619615639,
+      "grad_norm": 3.3458366768646193,
+      "learning_rate": 9.9805150892157e-06,
+      "loss": 0.25424268,
+      "memory(GiB)": 33.01,
+      "step": 4630,
+      "train_speed(iter/s)": 0.188635
+    },
+    {
+      "acc": 0.9459053,
+      "epoch": 1.5357852882703777,
+      "grad_norm": 2.2689321431155234,
+      "learning_rate": 9.980394106806639e-06,
+      "loss": 0.25565271,
+      "memory(GiB)": 33.01,
+      "step": 4635,
+      "train_speed(iter/s)": 0.188639
+    },
+    {
+      "acc": 0.93355293,
+      "epoch": 1.5374420145791916,
+      "grad_norm": 2.7753707714720517,
+      "learning_rate": 9.980272750705245e-06,
+      "loss": 0.28458288,
+      "memory(GiB)": 33.01,
+      "step": 4640,
+      "train_speed(iter/s)": 0.188644
+    },
+    {
+      "acc": 0.9476614,
+      "epoch": 1.5390987408880052,
+      "grad_norm": 2.3229287601639133,
+      "learning_rate": 9.980151020920627e-06,
+      "loss": 0.25084689,
+      "memory(GiB)": 33.01,
+      "step": 4645,
+      "train_speed(iter/s)": 0.188649
+    },
+    {
+      "acc": 0.95184021,
+      "epoch": 1.540755467196819,
+      "grad_norm": 6.83150151463161,
+      "learning_rate": 9.980028917461918e-06,
+      "loss": 0.24106314,
+      "memory(GiB)": 33.01,
+      "step": 4650,
+      "train_speed(iter/s)": 0.188653
+    },
+    {
+      "acc": 0.9499938,
+      "epoch": 1.542412193505633,
+      "grad_norm": 4.508910679339447,
+      "learning_rate": 9.979906440338279e-06,
+      "loss": 0.2259788,
+      "memory(GiB)": 33.01,
+      "step": 4655,
+      "train_speed(iter/s)": 0.188658
+    },
+    {
+      "acc": 0.94400835,
+      "epoch": 1.5440689198144466,
+      "grad_norm": 4.486912101753494,
+      "learning_rate": 9.979783589558905e-06,
+      "loss": 0.2795326,
+      "memory(GiB)": 33.01,
+      "step": 4660,
+      "train_speed(iter/s)": 0.188662
+    },
+    {
+      "acc": 0.93964567,
+      "epoch": 1.5457256461232605,
+      "grad_norm": 4.1657848852788355,
+      "learning_rate": 9.979660365133011e-06,
+      "loss": 0.26774707,
+      "memory(GiB)": 33.01,
+      "step": 4665,
+      "train_speed(iter/s)": 0.188666
+    },
+    {
+      "acc": 0.94199753,
+      "epoch": 1.5473823724320743,
+      "grad_norm": 2.7648572377320053,
+      "learning_rate": 9.979536767069845e-06,
+      "loss": 0.25348239,
+      "memory(GiB)": 33.01,
+      "step": 4670,
+      "train_speed(iter/s)": 0.188671
+    },
+    {
+      "acc": 0.95243158,
+      "epoch": 1.549039098740888,
+      "grad_norm": 2.8747702505039254,
+      "learning_rate": 9.979412795378682e-06,
+      "loss": 0.220208,
+      "memory(GiB)": 33.01,
+      "step": 4675,
+      "train_speed(iter/s)": 0.188671
+    },
+    {
+      "acc": 0.94590988,
+      "epoch": 1.5506958250497018,
+      "grad_norm": 5.076808549466851,
+      "learning_rate": 9.979288450068824e-06,
+      "loss": 0.23557816,
+      "memory(GiB)": 33.01,
+      "step": 4680,
+      "train_speed(iter/s)": 0.188672
+    },
+    {
+      "acc": 0.93664322,
+      "epoch": 1.5523525513585157,
+      "grad_norm": 4.425463261157371,
+      "learning_rate": 9.979163731149604e-06,
+      "loss": 0.31336203,
+      "memory(GiB)": 33.01,
+      "step": 4685,
+      "train_speed(iter/s)": 0.188668
+    },
+    {
+      "acc": 0.9431179,
+      "epoch": 1.5540092776673293,
+      "grad_norm": 3.126294777396309,
+      "learning_rate": 9.979038638630377e-06,
+      "loss": 0.294207,
+      "memory(GiB)": 33.01,
+      "step": 4690,
+      "train_speed(iter/s)": 0.188673
+    },
+    {
+      "acc": 0.95931683,
+      "epoch": 1.5556660039761432,
+      "grad_norm": 2.141774558741648,
+      "learning_rate": 9.978913172520536e-06,
+      "loss": 0.21437566,
+      "memory(GiB)": 33.01,
+      "step": 4695,
+      "train_speed(iter/s)": 0.188678
+    },
+    {
+      "acc": 0.94850693,
+      "epoch": 1.557322730284957,
+      "grad_norm": 2.2059357337336234,
+      "learning_rate": 9.97878733282949e-06,
+      "loss": 0.28407876,
+      "memory(GiB)": 33.01,
+      "step": 4700,
+      "train_speed(iter/s)": 0.188683
+    },
+    {
+      "acc": 0.95758886,
+      "epoch": 1.5589794565937707,
+      "grad_norm": 2.4757581341795243,
+      "learning_rate": 9.978661119566686e-06,
+      "loss": 0.19829302,
+      "memory(GiB)": 33.01,
+      "step": 4705,
+      "train_speed(iter/s)": 0.188687
+    },
+    {
+      "acc": 0.95511742,
+      "epoch": 1.5606361829025845,
+      "grad_norm": 3.0990654451934745,
+      "learning_rate": 9.978534532741593e-06,
+      "loss": 0.23894968,
+      "memory(GiB)": 33.01,
+      "step": 4710,
+      "train_speed(iter/s)": 0.188692
+    },
+    {
+      "acc": 0.9526618,
+      "epoch": 1.5622929092113984,
+      "grad_norm": 3.657655528211022,
+      "learning_rate": 9.97840757236371e-06,
+      "loss": 0.24336443,
+      "memory(GiB)": 33.01,
+      "step": 4715,
+      "train_speed(iter/s)": 0.188697
+    },
+    {
+      "acc": 0.94776039,
+      "epoch": 1.563949635520212,
+      "grad_norm": 5.621035773137675,
+      "learning_rate": 9.978280238442567e-06,
+      "loss": 0.23600438,
+      "memory(GiB)": 33.01,
+      "step": 4720,
+      "train_speed(iter/s)": 0.188701
+    },
+    {
+      "acc": 0.94383068,
+      "epoch": 1.5656063618290257,
+      "grad_norm": 4.32774043670886,
+      "learning_rate": 9.978152530987715e-06,
+      "loss": 0.24449029,
+      "memory(GiB)": 33.01,
+      "step": 4725,
+      "train_speed(iter/s)": 0.188706
+    },
+    {
+      "acc": 0.94421568,
+      "epoch": 1.5672630881378398,
+      "grad_norm": 3.33296997831563,
+      "learning_rate": 9.97802445000874e-06,
+      "loss": 0.25713592,
+      "memory(GiB)": 33.01,
+      "step": 4730,
+      "train_speed(iter/s)": 0.18871
+    },
+    {
+      "acc": 0.95428514,
+      "epoch": 1.5689198144466534,
+      "grad_norm": 1.6532099651446057,
+      "learning_rate": 9.977895995515252e-06,
+      "loss": 0.21798651,
+      "memory(GiB)": 33.01,
+      "step": 4735,
+      "train_speed(iter/s)": 0.188714
+    },
+    {
+      "acc": 0.9493392,
+      "epoch": 1.570576540755467,
+      "grad_norm": 3.740700380404861,
+      "learning_rate": 9.977767167516892e-06,
+      "loss": 0.23871698,
+      "memory(GiB)": 33.01,
+      "step": 4740,
+      "train_speed(iter/s)": 0.188719
+    },
+    {
+      "acc": 0.94671459,
+      "epoch": 1.5722332670642811,
+      "grad_norm": 3.368393810253313,
+      "learning_rate": 9.977637966023325e-06,
+      "loss": 0.23216662,
+      "memory(GiB)": 33.01,
+      "step": 4745,
+      "train_speed(iter/s)": 0.188723
+    },
+    {
+      "acc": 0.95063591,
+      "epoch": 1.5738899933730948,
+      "grad_norm": 3.0993943871932976,
+      "learning_rate": 9.977508391044248e-06,
+      "loss": 0.26406002,
+      "memory(GiB)": 33.01,
+      "step": 4750,
+      "train_speed(iter/s)": 0.188728
+    },
+    {
+      "acc": 0.94728756,
+      "epoch": 1.5755467196819084,
+      "grad_norm": 3.3800826559542623,
+      "learning_rate": 9.977378442589384e-06,
+      "loss": 0.2560663,
+      "memory(GiB)": 33.01,
+      "step": 4755,
+      "train_speed(iter/s)": 0.188732
+    },
+    {
+      "acc": 0.94631729,
+      "epoch": 1.5772034459907225,
+      "grad_norm": 3.327726538205645,
+      "learning_rate": 9.977248120668483e-06,
+      "loss": 0.24505229,
+      "memory(GiB)": 33.01,
+      "step": 4760,
+      "train_speed(iter/s)": 0.188737
+    },
+    {
+      "acc": 0.9512085,
+      "epoch": 1.5788601722995361,
+      "grad_norm": 4.054660356081657,
+      "learning_rate": 9.977117425291326e-06,
+      "loss": 0.25066209,
+      "memory(GiB)": 33.01,
+      "step": 4765,
+      "train_speed(iter/s)": 0.188742
+    },
+    {
+      "acc": 0.94238873,
+      "epoch": 1.5805168986083498,
+      "grad_norm": 2.68219292620574,
+      "learning_rate": 9.97698635646772e-06,
+      "loss": 0.24627228,
+      "memory(GiB)": 33.01,
+      "step": 4770,
+      "train_speed(iter/s)": 0.188746
+    },
+    {
+      "acc": 0.9424408,
+      "epoch": 1.5821736249171638,
+      "grad_norm": 3.6554527332658666,
+      "learning_rate": 9.976854914207498e-06,
+      "loss": 0.27461429,
+      "memory(GiB)": 33.01,
+      "step": 4775,
+      "train_speed(iter/s)": 0.18875
+    },
+    {
+      "acc": 0.9430851,
+      "epoch": 1.5838303512259775,
+      "grad_norm": 3.8964501710851858,
+      "learning_rate": 9.976723098520528e-06,
+      "loss": 0.26223986,
+      "memory(GiB)": 33.01,
+      "step": 4780,
+      "train_speed(iter/s)": 0.188755
+    },
+    {
+      "acc": 0.94343033,
+      "epoch": 1.5854870775347911,
+      "grad_norm": 3.369986079845925,
+      "learning_rate": 9.976590909416698e-06,
+      "loss": 0.26183143,
+      "memory(GiB)": 33.01,
+      "step": 4785,
+      "train_speed(iter/s)": 0.18876
+    },
+    {
+      "acc": 0.96031656,
+      "epoch": 1.587143803843605,
+      "grad_norm": 2.380405076752628,
+      "learning_rate": 9.97645834690593e-06,
+      "loss": 0.19646028,
+      "memory(GiB)": 33.01,
+      "step": 4790,
+      "train_speed(iter/s)": 0.188764
+    },
+    {
+      "acc": 0.95498562,
+      "epoch": 1.5888005301524188,
+      "grad_norm": 4.297364822205282,
+      "learning_rate": 9.976325410998168e-06,
+      "loss": 0.22078636,
+      "memory(GiB)": 33.01,
+      "step": 4795,
+      "train_speed(iter/s)": 0.188768
+    },
+    {
+      "acc": 0.94994173,
+      "epoch": 1.5904572564612325,
+      "grad_norm": 2.547088024508513,
+      "learning_rate": 9.976192101703393e-06,
+      "loss": 0.23789809,
+      "memory(GiB)": 33.01,
+      "step": 4800,
+      "train_speed(iter/s)": 0.188773
+    },
+    {
+      "acc": 0.95071087,
+      "epoch": 1.5921139827700463,
+      "grad_norm": 2.895434807469631,
+      "learning_rate": 9.976058419031603e-06,
+      "loss": 0.24055889,
+      "memory(GiB)": 33.01,
+      "step": 4805,
+      "train_speed(iter/s)": 0.188778
+    },
+    {
+      "acc": 0.94927464,
+      "epoch": 1.5937707090788602,
+      "grad_norm": 2.7022893151934295,
+      "learning_rate": 9.975924362992832e-06,
+      "loss": 0.240098,
+      "memory(GiB)": 33.01,
+      "step": 4810,
+      "train_speed(iter/s)": 0.188782
+    },
+    {
+      "acc": 0.94955883,
+      "epoch": 1.5954274353876738,
+      "grad_norm": 3.098667505578039,
+      "learning_rate": 9.975789933597138e-06,
+      "loss": 0.29886706,
+      "memory(GiB)": 33.01,
+      "step": 4815,
+      "train_speed(iter/s)": 0.188786
+    },
+    {
+      "acc": 0.9490921,
+      "epoch": 1.5970841616964877,
+      "grad_norm": 3.1763937338831485,
+      "learning_rate": 9.975655130854612e-06,
+      "loss": 0.25676432,
+      "memory(GiB)": 33.01,
+      "step": 4820,
+      "train_speed(iter/s)": 0.188791
+    },
+    {
+      "acc": 0.94501534,
+      "epoch": 1.5987408880053016,
+      "grad_norm": 3.3037021375603874,
+      "learning_rate": 9.975519954775369e-06,
+      "loss": 0.26909356,
+      "memory(GiB)": 33.01,
+      "step": 4825,
+      "train_speed(iter/s)": 0.188795
+    },
+    {
+      "acc": 0.93495655,
+      "epoch": 1.6003976143141152,
+      "grad_norm": 5.088113469771529,
+      "learning_rate": 9.975384405369549e-06,
+      "loss": 0.27787559,
+      "memory(GiB)": 33.01,
+      "step": 4830,
+      "train_speed(iter/s)": 0.188799
+    },
+    {
+      "acc": 0.94772968,
+      "epoch": 1.602054340622929,
+      "grad_norm": 4.514764624259456,
+      "learning_rate": 9.975248482647327e-06,
+      "loss": 0.23549538,
+      "memory(GiB)": 33.01,
+      "step": 4835,
+      "train_speed(iter/s)": 0.188804
+    },
+    {
+      "acc": 0.9459712,
+      "epoch": 1.603711066931743,
+      "grad_norm": 2.877098089780718,
+      "learning_rate": 9.975112186618902e-06,
+      "loss": 0.27602563,
+      "memory(GiB)": 33.01,
+      "step": 4840,
+      "train_speed(iter/s)": 0.188808
+    },
+    {
+      "acc": 0.96013279,
+      "epoch": 1.6053677932405566,
+      "grad_norm": 2.974592013638567,
+      "learning_rate": 9.9749755172945e-06,
+      "loss": 0.2202702,
+      "memory(GiB)": 33.01,
+      "step": 4845,
+      "train_speed(iter/s)": 0.188812
+    },
+    {
+      "acc": 0.96109982,
+      "epoch": 1.6070245195493704,
+      "grad_norm": 3.0350816349287246,
+      "learning_rate": 9.974838474684379e-06,
+      "loss": 0.1902697,
+      "memory(GiB)": 33.01,
+      "step": 4850,
+      "train_speed(iter/s)": 0.188816
+    },
+    {
+      "acc": 0.95083942,
+      "epoch": 1.6086812458581843,
+      "grad_norm": 6.679294419395035,
+      "learning_rate": 9.974701058798822e-06,
+      "loss": 0.23198597,
+      "memory(GiB)": 33.01,
+      "step": 4855,
+      "train_speed(iter/s)": 0.18882
+    },
+    {
+      "acc": 0.94768581,
+      "epoch": 1.610337972166998,
+      "grad_norm": 3.9727156126425944,
+      "learning_rate": 9.97456326964814e-06,
+      "loss": 0.23015182,
+      "memory(GiB)": 33.01,
+      "step": 4860,
+      "train_speed(iter/s)": 0.188824
+    },
+    {
+      "acc": 0.95107021,
+      "epoch": 1.6119946984758118,
+      "grad_norm": 4.535114055899051,
+      "learning_rate": 9.974425107242673e-06,
+      "loss": 0.23106141,
+      "memory(GiB)": 33.01,
+      "step": 4865,
+      "train_speed(iter/s)": 0.188829
+    },
+    {
+      "acc": 0.9494894,
+      "epoch": 1.6136514247846256,
+      "grad_norm": 3.870704983276642,
+      "learning_rate": 9.974286571592788e-06,
+      "loss": 0.26235733,
+      "memory(GiB)": 33.01,
+      "step": 4870,
+      "train_speed(iter/s)": 0.188833
+    },
+    {
+      "acc": 0.94805145,
+      "epoch": 1.6153081510934393,
+      "grad_norm": 3.3742441210245797,
+      "learning_rate": 9.974147662708883e-06,
+      "loss": 0.21995747,
+      "memory(GiB)": 33.01,
+      "step": 4875,
+      "train_speed(iter/s)": 0.188837
+    },
+    {
+      "acc": 0.94874687,
+      "epoch": 1.6169648774022531,
+      "grad_norm": 3.401021821485584,
+      "learning_rate": 9.974008380601382e-06,
+      "loss": 0.24643075,
+      "memory(GiB)": 33.01,
+      "step": 4880,
+      "train_speed(iter/s)": 0.188842
+    },
+    {
+      "acc": 0.95776138,
+      "epoch": 1.618621603711067,
+      "grad_norm": 2.865593808145345,
+      "learning_rate": 9.973868725280732e-06,
+      "loss": 0.1929089,
+      "memory(GiB)": 33.01,
+      "step": 4885,
+      "train_speed(iter/s)": 0.188846
+    },
+    {
+      "acc": 0.94048004,
+      "epoch": 1.6202783300198806,
+      "grad_norm": 3.623357009895858,
+      "learning_rate": 9.973728696757416e-06,
+      "loss": 0.26675167,
+      "memory(GiB)": 33.01,
+      "step": 4890,
+      "train_speed(iter/s)": 0.188851
+    },
+    {
+      "acc": 0.95277262,
+      "epoch": 1.6219350563286945,
+      "grad_norm": 4.57397452194311,
+      "learning_rate": 9.973588295041944e-06,
+      "loss": 0.22158756,
+      "memory(GiB)": 33.01,
+      "step": 4895,
+      "train_speed(iter/s)": 0.188855
+    },
+    {
+      "acc": 0.9544117,
+      "epoch": 1.6235917826375084,
+      "grad_norm": 3.77915181580508,
+      "learning_rate": 9.973447520144849e-06,
+      "loss": 0.23694935,
+      "memory(GiB)": 33.01,
+      "step": 4900,
+      "train_speed(iter/s)": 0.188859
+    },
+    {
+      "acc": 0.95348244,
+      "epoch": 1.625248508946322,
+      "grad_norm": 3.0287842565991356,
+      "learning_rate": 9.973306372076694e-06,
+      "loss": 0.2032382,
+      "memory(GiB)": 33.01,
+      "step": 4905,
+      "train_speed(iter/s)": 0.188859
+    },
+    {
+      "acc": 0.94308414,
+      "epoch": 1.6269052352551359,
+      "grad_norm": 2.7091908324528453,
+      "learning_rate": 9.973164850848075e-06,
+      "loss": 0.25078931,
+      "memory(GiB)": 33.01,
+      "step": 4910,
+      "train_speed(iter/s)": 0.18886
+    },
+    {
+      "acc": 0.94056168,
+      "epoch": 1.6285619615639497,
+      "grad_norm": 4.888642428157676,
+      "learning_rate": 9.973022956469607e-06,
+      "loss": 0.26888967,
+      "memory(GiB)": 33.01,
+      "step": 4915,
+      "train_speed(iter/s)": 0.188857
+    },
+    {
+      "acc": 0.94573936,
+      "epoch": 1.6302186878727634,
+      "grad_norm": 3.0506240931691915,
+      "learning_rate": 9.972880688951938e-06,
+      "loss": 0.2353272,
+      "memory(GiB)": 33.01,
+      "step": 4920,
+      "train_speed(iter/s)": 0.188862
+    },
+    {
+      "acc": 0.95126534,
+      "epoch": 1.6318754141815772,
+      "grad_norm": 3.418758115114436,
+      "learning_rate": 9.972738048305746e-06,
+      "loss": 0.23389206,
+      "memory(GiB)": 33.01,
+      "step": 4925,
+      "train_speed(iter/s)": 0.188866
+    },
+    {
+      "acc": 0.9515564,
+      "epoch": 1.633532140490391,
+      "grad_norm": 3.95230001541203,
+      "learning_rate": 9.972595034541736e-06,
+      "loss": 0.22550275,
+      "memory(GiB)": 33.01,
+      "step": 4930,
+      "train_speed(iter/s)": 0.18887
+    },
+    {
+      "acc": 0.95114269,
+      "epoch": 1.6351888667992047,
+      "grad_norm": 2.7403865006896124,
+      "learning_rate": 9.972451647670637e-06,
+      "loss": 0.21844039,
+      "memory(GiB)": 33.01,
+      "step": 4935,
+      "train_speed(iter/s)": 0.188875
+    },
+    {
+      "acc": 0.94797792,
+      "epoch": 1.6368455931080186,
+      "grad_norm": 3.7388123912333575,
+      "learning_rate": 9.972307887703209e-06,
+      "loss": 0.24305737,
+      "memory(GiB)": 33.01,
+      "step": 4940,
+      "train_speed(iter/s)": 0.188879
+    },
+    {
+      "acc": 0.95092726,
+      "epoch": 1.6385023194168324,
+      "grad_norm": 3.118229751230935,
+      "learning_rate": 9.972163754650241e-06,
+      "loss": 0.22735724,
+      "memory(GiB)": 33.01,
+      "step": 4945,
+      "train_speed(iter/s)": 0.188884
+    },
+    {
+      "acc": 0.95121222,
+      "epoch": 1.640159045725646,
+      "grad_norm": 3.5823478676548786,
+      "learning_rate": 9.972019248522549e-06,
+      "loss": 0.21556726,
+      "memory(GiB)": 33.01,
+      "step": 4950,
+      "train_speed(iter/s)": 0.188888
+    },
+    {
+      "acc": 0.94801369,
+      "epoch": 1.64181577203446,
+      "grad_norm": 3.5514396644121535,
+      "learning_rate": 9.971874369330977e-06,
+      "loss": 0.24600191,
+      "memory(GiB)": 33.01,
+      "step": 4955,
+      "train_speed(iter/s)": 0.188893
+    },
+    {
+      "acc": 0.95595074,
+      "epoch": 1.6434724983432738,
+      "grad_norm": 2.804197111880328,
+      "learning_rate": 9.971729117086393e-06,
+      "loss": 0.21580234,
+      "memory(GiB)": 33.01,
+      "step": 4960,
+      "train_speed(iter/s)": 0.188898
+    },
+    {
+      "acc": 0.95798407,
+      "epoch": 1.6451292246520874,
+      "grad_norm": 4.845645575346084,
+      "learning_rate": 9.971583491799704e-06,
+      "loss": 0.20477767,
+      "memory(GiB)": 33.01,
+      "step": 4965,
+      "train_speed(iter/s)": 0.188902
+    },
+    {
+      "acc": 0.94212322,
+      "epoch": 1.6467859509609013,
+      "grad_norm": 4.103328640046485,
+      "learning_rate": 9.971437493481827e-06,
+      "loss": 0.23042445,
+      "memory(GiB)": 33.01,
+      "step": 4970,
+      "train_speed(iter/s)": 0.188906
+    },
+    {
+      "acc": 0.96105289,
+      "epoch": 1.6484426772697152,
+      "grad_norm": 2.7529876836341582,
+      "learning_rate": 9.97129112214373e-06,
+      "loss": 0.20002341,
+      "memory(GiB)": 33.01,
+      "step": 4975,
+      "train_speed(iter/s)": 0.18891
+    },
+    {
+      "acc": 0.94937401,
+      "epoch": 1.6500994035785288,
+      "grad_norm": 3.563674862874002,
+      "learning_rate": 9.971144377796387e-06,
+      "loss": 0.21998301,
+      "memory(GiB)": 33.01,
+      "step": 4980,
+      "train_speed(iter/s)": 0.188915
+    },
+    {
+      "acc": 0.9542902,
+      "epoch": 1.6517561298873427,
+      "grad_norm": 4.074266167183439,
+      "learning_rate": 9.970997260450817e-06,
+      "loss": 0.20854292,
+      "memory(GiB)": 33.01,
+      "step": 4985,
+      "train_speed(iter/s)": 0.188919
+    },
+    {
+      "acc": 0.96340275,
+      "epoch": 1.6534128561961565,
+      "grad_norm": 2.5160689305384083,
+      "learning_rate": 9.970849770118053e-06,
+      "loss": 0.18839086,
+      "memory(GiB)": 33.01,
+      "step": 4990,
+      "train_speed(iter/s)": 0.188923
+    },
+    {
+      "acc": 0.95943623,
+      "epoch": 1.6550695825049702,
+      "grad_norm": 3.4100784753401157,
+      "learning_rate": 9.970701906809169e-06,
+      "loss": 0.20573511,
+      "memory(GiB)": 33.01,
+      "step": 4995,
+      "train_speed(iter/s)": 0.188927
+    },
+    {
+      "acc": 0.95086899,
+      "epoch": 1.6567263088137838,
+      "grad_norm": 3.6252402944198403,
+      "learning_rate": 9.970553670535254e-06,
+      "loss": 0.23412619,
+      "memory(GiB)": 33.01,
+      "step": 5000,
+      "train_speed(iter/s)": 0.188931
+    },
+    {
+      "acc": 0.94626532,
+      "epoch": 1.6583830351225979,
+      "grad_norm": 3.5790919703147615,
+      "learning_rate": 9.970405061307439e-06,
+      "loss": 0.25025907,
+      "memory(GiB)": 33.01,
+      "step": 5005,
+      "train_speed(iter/s)": 0.188936
+    },
+    {
+      "acc": 0.94169321,
+      "epoch": 1.6600397614314115,
+      "grad_norm": 4.397053805409449,
+      "learning_rate": 9.97025607913687e-06,
+      "loss": 0.25981188,
+      "memory(GiB)": 33.01,
+      "step": 5010,
+      "train_speed(iter/s)": 0.18894
+    },
+    {
+      "acc": 0.95897369,
+      "epoch": 1.6616964877402252,
+      "grad_norm": 3.873323476185693,
+      "learning_rate": 9.97010672403473e-06,
+      "loss": 0.1975831,
+      "memory(GiB)": 33.01,
+      "step": 5015,
+      "train_speed(iter/s)": 0.188944
+    },
+    {
+      "acc": 0.94391537,
+      "epoch": 1.6633532140490392,
+      "grad_norm": 3.702807900229647,
+      "learning_rate": 9.969956996012225e-06,
+      "loss": 0.25931425,
+      "memory(GiB)": 33.01,
+      "step": 5020,
+      "train_speed(iter/s)": 0.188948
+    },
+    {
+      "acc": 0.9463459,
+      "epoch": 1.6650099403578529,
+      "grad_norm": 2.7456736700212057,
+      "learning_rate": 9.96980689508059e-06,
+      "loss": 0.22293758,
+      "memory(GiB)": 33.01,
+      "step": 5025,
+      "train_speed(iter/s)": 0.188953
+    },
+    {
+      "acc": 0.95056171,
+      "epoch": 1.6666666666666665,
+      "grad_norm": 4.5307337977824504,
+      "learning_rate": 9.96965642125109e-06,
+      "loss": 0.24241471,
+      "memory(GiB)": 33.01,
+      "step": 5030,
+      "train_speed(iter/s)": 0.188957
+    },
+    {
+      "acc": 0.95831909,
+      "epoch": 1.6683233929754806,
+      "grad_norm": 2.5561137922719492,
+      "learning_rate": 9.969505574535015e-06,
+      "loss": 0.21591077,
+      "memory(GiB)": 33.01,
+      "step": 5035,
+      "train_speed(iter/s)": 0.188962
+    },
+    {
+      "acc": 0.94991932,
+      "epoch": 1.6699801192842942,
+      "grad_norm": 3.988832293147868,
+      "learning_rate": 9.969354354943687e-06,
+      "loss": 0.25156703,
+      "memory(GiB)": 33.01,
+      "step": 5040,
+      "train_speed(iter/s)": 0.188966
+    },
+    {
+      "acc": 0.94121628,
+      "epoch": 1.6716368455931079,
+      "grad_norm": 3.322732560737147,
+      "learning_rate": 9.969202762488452e-06,
+      "loss": 0.27342329,
+      "memory(GiB)": 33.01,
+      "step": 5045,
+      "train_speed(iter/s)": 0.18897
+    },
+    {
+      "acc": 0.94726105,
+      "epoch": 1.673293571901922,
+      "grad_norm": 3.1076731504376363,
+      "learning_rate": 9.969050797180687e-06,
+      "loss": 0.228824,
+      "memory(GiB)": 33.01,
+      "step": 5050,
+      "train_speed(iter/s)": 0.188975
+    },
+    {
+      "acc": 0.9406065,
+      "epoch": 1.6749502982107356,
+      "grad_norm": 2.715517351499893,
+      "learning_rate": 9.968898459031793e-06,
+      "loss": 0.22380838,
+      "memory(GiB)": 33.01,
+      "step": 5055,
+      "train_speed(iter/s)": 0.188979
+    },
+    {
+      "acc": 0.95967426,
+      "epoch": 1.6766070245195492,
+      "grad_norm": 2.5665196685724334,
+      "learning_rate": 9.968745748053203e-06,
+      "loss": 0.2053267,
+      "memory(GiB)": 33.01,
+      "step": 5060,
+      "train_speed(iter/s)": 0.188984
+    },
+    {
+      "acc": 0.95684433,
+      "epoch": 1.678263750828363,
+      "grad_norm": 3.125254186816331,
+      "learning_rate": 9.968592664256378e-06,
+      "loss": 0.21209474,
+      "memory(GiB)": 33.01,
+      "step": 5065,
+      "train_speed(iter/s)": 0.188988
+    },
+    {
+      "acc": 0.95527973,
+      "epoch": 1.679920477137177,
+      "grad_norm": 4.22308090489758,
+      "learning_rate": 9.968439207652801e-06,
+      "loss": 0.21446476,
+      "memory(GiB)": 33.01,
+      "step": 5070,
+      "train_speed(iter/s)": 0.188992
+    },
+    {
+      "acc": 0.95776043,
+      "epoch": 1.6815772034459906,
+      "grad_norm": 3.2310268009096,
+      "learning_rate": 9.96828537825399e-06,
+      "loss": 0.20503042,
+      "memory(GiB)": 33.01,
+      "step": 5075,
+      "train_speed(iter/s)": 0.188996
+    },
+    {
+      "acc": 0.95915852,
+      "epoch": 1.6832339297548045,
+      "grad_norm": 3.8335144505767587,
+      "learning_rate": 9.96813117607149e-06,
+      "loss": 0.17421991,
+      "memory(GiB)": 33.01,
+      "step": 5080,
+      "train_speed(iter/s)": 0.189
+    },
+    {
+      "acc": 0.9563695,
+      "epoch": 1.6848906560636183,
+      "grad_norm": 2.8695194670299995,
+      "learning_rate": 9.967976601116873e-06,
+      "loss": 0.21266487,
+      "memory(GiB)": 33.01,
+      "step": 5085,
+      "train_speed(iter/s)": 0.189004
+    },
+    {
+      "acc": 0.96290131,
+      "epoch": 1.686547382372432,
+      "grad_norm": 4.501683510959438,
+      "learning_rate": 9.967821653401734e-06,
+      "loss": 0.16936184,
+      "memory(GiB)": 33.01,
+      "step": 5090,
+      "train_speed(iter/s)": 0.189008
+    },
+    {
+      "acc": 0.95391636,
+      "epoch": 1.6882041086812458,
+      "grad_norm": 4.418605474638244,
+      "learning_rate": 9.967666332937702e-06,
+      "loss": 0.23554609,
+      "memory(GiB)": 33.01,
+      "step": 5095,
+      "train_speed(iter/s)": 0.189012
+    },
+    {
+      "acc": 0.9614645,
+      "epoch": 1.6898608349900597,
+      "grad_norm": 3.942685151596654,
+      "learning_rate": 9.967510639736432e-06,
+      "loss": 0.23986785,
+      "memory(GiB)": 33.01,
+      "step": 5100,
+      "train_speed(iter/s)": 0.189016
+    },
+    {
+      "acc": 0.95797901,
+      "epoch": 1.6915175612988733,
+      "grad_norm": 4.000520180033358,
+      "learning_rate": 9.967354573809612e-06,
+      "loss": 0.20662632,
+      "memory(GiB)": 33.01,
+      "step": 5105,
+      "train_speed(iter/s)": 0.18902
+    },
+    {
+      "acc": 0.9563776,
+      "epoch": 1.6931742876076872,
+      "grad_norm": 3.002694099791044,
+      "learning_rate": 9.967198135168945e-06,
+      "loss": 0.21235328,
+      "memory(GiB)": 33.01,
+      "step": 5110,
+      "train_speed(iter/s)": 0.189025
+    },
+    {
+      "acc": 0.94639397,
+      "epoch": 1.694831013916501,
+      "grad_norm": 3.3130267697232045,
+      "learning_rate": 9.967041323826176e-06,
+      "loss": 0.25527148,
+      "memory(GiB)": 33.01,
+      "step": 5115,
+      "train_speed(iter/s)": 0.189029
+    },
+    {
+      "acc": 0.95592213,
+      "epoch": 1.6964877402253147,
+      "grad_norm": 3.336906608094463,
+      "learning_rate": 9.96688413979307e-06,
+      "loss": 0.20270851,
+      "memory(GiB)": 33.01,
+      "step": 5120,
+      "train_speed(iter/s)": 0.189033
+    },
+    {
+      "acc": 0.95745296,
+      "epoch": 1.6981444665341285,
+      "grad_norm": 4.44821439755114,
+      "learning_rate": 9.966726583081425e-06,
+      "loss": 0.20269413,
+      "memory(GiB)": 33.01,
+      "step": 5125,
+      "train_speed(iter/s)": 0.189037
+    },
+    {
+      "acc": 0.94841909,
+      "epoch": 1.6998011928429424,
+      "grad_norm": 3.580350601322894,
+      "learning_rate": 9.966568653703059e-06,
+      "loss": 0.2257447,
+      "memory(GiB)": 33.01,
+      "step": 5130,
+      "train_speed(iter/s)": 0.189041
+    },
+    {
+      "acc": 0.95071068,
+      "epoch": 1.701457919151756,
+      "grad_norm": 3.428909422067706,
+      "learning_rate": 9.966410351669827e-06,
+      "loss": 0.23408895,
+      "memory(GiB)": 33.01,
+      "step": 5135,
+      "train_speed(iter/s)": 0.189039
+    },
+    {
+      "acc": 0.964328,
+      "epoch": 1.70311464546057,
+      "grad_norm": 4.208990235126206,
+      "learning_rate": 9.96625167699361e-06,
+      "loss": 0.17427733,
+      "memory(GiB)": 33.01,
+      "step": 5140,
+      "train_speed(iter/s)": 0.189039
+    },
+    {
+      "acc": 0.95825062,
+      "epoch": 1.7047713717693838,
+      "grad_norm": 3.353043995134885,
+      "learning_rate": 9.966092629686306e-06,
+      "loss": 0.18491567,
+      "memory(GiB)": 33.01,
+      "step": 5145,
+      "train_speed(iter/s)": 0.189039
+    },
+    {
+      "acc": 0.95702209,
+      "epoch": 1.7064280980781974,
+      "grad_norm": 2.9130928584040086,
+      "learning_rate": 9.96593320975986e-06,
+      "loss": 0.19480469,
+      "memory(GiB)": 33.01,
+      "step": 5150,
+      "train_speed(iter/s)": 0.189044
+    },
+    {
+      "acc": 0.94524212,
+      "epoch": 1.7080848243870113,
+      "grad_norm": 3.237677337789052,
+      "learning_rate": 9.96577341722623e-06,
+      "loss": 0.21276062,
+      "memory(GiB)": 33.01,
+      "step": 5155,
+      "train_speed(iter/s)": 0.189048
+    },
+    {
+      "acc": 0.95643759,
+      "epoch": 1.7097415506958251,
+      "grad_norm": 3.0503011618136227,
+      "learning_rate": 9.965613252097407e-06,
+      "loss": 0.22177296,
+      "memory(GiB)": 33.01,
+      "step": 5160,
+      "train_speed(iter/s)": 0.189052
+    },
+    {
+      "acc": 0.95698528,
+      "epoch": 1.7113982770046388,
+      "grad_norm": 3.383540295767972,
+      "learning_rate": 9.965452714385411e-06,
+      "loss": 0.1991153,
+      "memory(GiB)": 33.01,
+      "step": 5165,
+      "train_speed(iter/s)": 0.189056
+    },
+    {
+      "acc": 0.96373568,
+      "epoch": 1.7130550033134526,
+      "grad_norm": 3.5158197989772724,
+      "learning_rate": 9.965291804102286e-06,
+      "loss": 0.20207872,
+      "memory(GiB)": 33.01,
+      "step": 5170,
+      "train_speed(iter/s)": 0.18906
+    },
+    {
+      "acc": 0.95304537,
+      "epoch": 1.7147117296222665,
+      "grad_norm": 3.345050508240883,
+      "learning_rate": 9.965130521260114e-06,
+      "loss": 0.22512956,
+      "memory(GiB)": 33.01,
+      "step": 5175,
+      "train_speed(iter/s)": 0.189064
+    },
+    {
+      "acc": 0.95467625,
+      "epoch": 1.7163684559310801,
+      "grad_norm": 3.7761870019739443,
+      "learning_rate": 9.964968865870988e-06,
+      "loss": 0.21822906,
+      "memory(GiB)": 33.01,
+      "step": 5180,
+      "train_speed(iter/s)": 0.189069
+    },
+    {
+      "acc": 0.95449238,
+      "epoch": 1.718025182239894,
+      "grad_norm": 6.041996880283651,
+      "learning_rate": 9.964806837947046e-06,
+      "loss": 0.21440597,
+      "memory(GiB)": 33.01,
+      "step": 5185,
+      "train_speed(iter/s)": 0.189073
+    },
+    {
+      "acc": 0.95477943,
+      "epoch": 1.7196819085487078,
+      "grad_norm": 3.384474313017865,
+      "learning_rate": 9.964644437500442e-06,
+      "loss": 0.19661164,
+      "memory(GiB)": 33.01,
+      "step": 5190,
+      "train_speed(iter/s)": 0.189077
+    },
+    {
+      "acc": 0.95581684,
+      "epoch": 1.7213386348575215,
+      "grad_norm": 2.7932712151654995,
+      "learning_rate": 9.964481664543367e-06,
+      "loss": 0.22425969,
+      "memory(GiB)": 33.01,
+      "step": 5195,
+      "train_speed(iter/s)": 0.189081
+    },
+    {
+      "acc": 0.95531349,
+      "epoch": 1.7229953611663353,
+      "grad_norm": 6.418297480498033,
+      "learning_rate": 9.964318519088031e-06,
+      "loss": 0.22553568,
+      "memory(GiB)": 33.01,
+      "step": 5200,
+      "train_speed(iter/s)": 0.189086
+    },
+    {
+      "acc": 0.95457287,
+      "epoch": 1.7246520874751492,
+      "grad_norm": 6.0263138539503665,
+      "learning_rate": 9.96415500114668e-06,
+      "loss": 0.20696769,
+      "memory(GiB)": 33.01,
+      "step": 5205,
+      "train_speed(iter/s)": 0.18909
+    },
+    {
+      "acc": 0.9520792,
+      "epoch": 1.7263088137839628,
+      "grad_norm": 3.8225435874716744,
+      "learning_rate": 9.96399111073158e-06,
+      "loss": 0.22758288,
+      "memory(GiB)": 33.01,
+      "step": 5210,
+      "train_speed(iter/s)": 0.189094
+    },
+    {
+      "acc": 0.96496429,
+      "epoch": 1.7279655400927767,
+      "grad_norm": 3.6492084490850134,
+      "learning_rate": 9.963826847855036e-06,
+      "loss": 0.19169132,
+      "memory(GiB)": 33.01,
+      "step": 5215,
+      "train_speed(iter/s)": 0.189098
+    },
+    {
+      "acc": 0.95372238,
+      "epoch": 1.7296222664015906,
+      "grad_norm": 5.643056842595738,
+      "learning_rate": 9.963662212529367e-06,
+      "loss": 0.22640302,
+      "memory(GiB)": 33.01,
+      "step": 5220,
+      "train_speed(iter/s)": 0.189102
+    },
+    {
+      "acc": 0.95633678,
+      "epoch": 1.7312789927104042,
+      "grad_norm": 6.03271100442074,
+      "learning_rate": 9.963497204766932e-06,
+      "loss": 0.21259847,
+      "memory(GiB)": 33.01,
+      "step": 5225,
+      "train_speed(iter/s)": 0.189107
+    },
+    {
+      "acc": 0.96206284,
+      "epoch": 1.732935719019218,
+      "grad_norm": 4.26004300569132,
+      "learning_rate": 9.963331824580114e-06,
+      "loss": 0.21050191,
+      "memory(GiB)": 33.01,
+      "step": 5230,
+      "train_speed(iter/s)": 0.189111
+    },
+    {
+      "acc": 0.957693,
+      "epoch": 1.734592445328032,
+      "grad_norm": 3.9193858249004085,
+      "learning_rate": 9.96316607198132e-06,
+      "loss": 0.20651617,
+      "memory(GiB)": 33.01,
+      "step": 5235,
+      "train_speed(iter/s)": 0.189115
+    },
+    {
+      "acc": 0.96739578,
+      "epoch": 1.7362491716368456,
+      "grad_norm": 2.4654973008604637,
+      "learning_rate": 9.962999946982989e-06,
+      "loss": 0.16179316,
+      "memory(GiB)": 33.01,
+      "step": 5240,
+      "train_speed(iter/s)": 0.18912
+    },
+    {
+      "acc": 0.96806068,
+      "epoch": 1.7379058979456594,
+      "grad_norm": 2.962843351851986,
+      "learning_rate": 9.962833449597588e-06,
+      "loss": 0.16479814,
+      "memory(GiB)": 33.01,
+      "step": 5245,
+      "train_speed(iter/s)": 0.189124
+    },
+    {
+      "acc": 0.95996628,
+      "epoch": 1.7395626242544733,
+      "grad_norm": 4.598241745371233,
+      "learning_rate": 9.962666579837611e-06,
+      "loss": 0.2062604,
+      "memory(GiB)": 33.01,
+      "step": 5250,
+      "train_speed(iter/s)": 0.189129
+    },
+    {
+      "acc": 0.96349878,
+      "epoch": 1.741219350563287,
+      "grad_norm": 3.3606613228712434,
+      "learning_rate": 9.962499337715579e-06,
+      "loss": 0.18981156,
+      "memory(GiB)": 33.01,
+      "step": 5255,
+      "train_speed(iter/s)": 0.189133
+    },
+    {
+      "acc": 0.94931068,
+      "epoch": 1.7428760768721008,
+      "grad_norm": 6.29495352456756,
+      "learning_rate": 9.962331723244042e-06,
+      "loss": 0.22791972,
+      "memory(GiB)": 33.01,
+      "step": 5260,
+      "train_speed(iter/s)": 0.189138
+    },
+    {
+      "acc": 0.95727224,
+      "epoch": 1.7445328031809146,
+      "grad_norm": 3.1276591507350804,
+      "learning_rate": 9.962163736435577e-06,
+      "loss": 0.21037591,
+      "memory(GiB)": 33.01,
+      "step": 5265,
+      "train_speed(iter/s)": 0.189142
+    },
+    {
+      "acc": 0.95165434,
+      "epoch": 1.7461895294897283,
+      "grad_norm": 3.2222867572456955,
+      "learning_rate": 9.961995377302792e-06,
+      "loss": 0.23067889,
+      "memory(GiB)": 33.01,
+      "step": 5270,
+      "train_speed(iter/s)": 0.189147
+    },
+    {
+      "acc": 0.9511673,
+      "epoch": 1.747846255798542,
+      "grad_norm": 2.621941939360323,
+      "learning_rate": 9.961826645858319e-06,
+      "loss": 0.21288648,
+      "memory(GiB)": 33.01,
+      "step": 5275,
+      "train_speed(iter/s)": 0.189151
+    },
+    {
+      "acc": 0.95167789,
+      "epoch": 1.749502982107356,
+      "grad_norm": 2.407501995466345,
+      "learning_rate": 9.961657542114822e-06,
+      "loss": 0.22208667,
+      "memory(GiB)": 33.01,
+      "step": 5280,
+      "train_speed(iter/s)": 0.189155
+    },
+    {
+      "acc": 0.94799423,
+      "epoch": 1.7511597084161696,
+      "grad_norm": 2.637185287430631,
+      "learning_rate": 9.961488066084985e-06,
+      "loss": 0.2063293,
+      "memory(GiB)": 33.01,
+      "step": 5285,
+      "train_speed(iter/s)": 0.18916
+    },
+    {
+      "acc": 0.96038609,
+      "epoch": 1.7528164347249833,
+      "grad_norm": 3.581585201701416,
+      "learning_rate": 9.961318217781532e-06,
+      "loss": 0.20009751,
+      "memory(GiB)": 33.01,
+      "step": 5290,
+      "train_speed(iter/s)": 0.189164
+    },
+    {
+      "acc": 0.95909309,
+      "epoch": 1.7544731610337974,
+      "grad_norm": 4.104222150516389,
+      "learning_rate": 9.961147997217207e-06,
+      "loss": 0.20937865,
+      "memory(GiB)": 33.01,
+      "step": 5295,
+      "train_speed(iter/s)": 0.189169
+    },
+    {
+      "acc": 0.96688213,
+      "epoch": 1.756129887342611,
+      "grad_norm": 2.178581712796887,
+      "learning_rate": 9.96097740440478e-06,
+      "loss": 0.17518711,
+      "memory(GiB)": 33.01,
+      "step": 5300,
+      "train_speed(iter/s)": 0.189173
+    },
+    {
+      "acc": 0.96675243,
+      "epoch": 1.7577866136514246,
+      "grad_norm": 5.061705003265265,
+      "learning_rate": 9.960806439357056e-06,
+      "loss": 0.18036057,
+      "memory(GiB)": 33.01,
+      "step": 5305,
+      "train_speed(iter/s)": 0.189177
+    },
+    {
+      "acc": 0.95813522,
+      "epoch": 1.7594433399602387,
+      "grad_norm": 3.2841932457130474,
+      "learning_rate": 9.96063510208686e-06,
+      "loss": 0.18793486,
+      "memory(GiB)": 33.01,
+      "step": 5310,
+      "train_speed(iter/s)": 0.189182
+    },
+    {
+      "acc": 0.96192608,
+      "epoch": 1.7611000662690524,
+      "grad_norm": 4.663571356022182,
+      "learning_rate": 9.960463392607053e-06,
+      "loss": 0.20335965,
+      "memory(GiB)": 33.01,
+      "step": 5315,
+      "train_speed(iter/s)": 0.189186
+    },
+    {
+      "acc": 0.96494179,
+      "epoch": 1.762756792577866,
+      "grad_norm": 3.5938288342932374,
+      "learning_rate": 9.960291310930522e-06,
+      "loss": 0.17716209,
+      "memory(GiB)": 33.01,
+      "step": 5320,
+      "train_speed(iter/s)": 0.189191
+    },
+    {
+      "acc": 0.95762358,
+      "epoch": 1.76441351888668,
+      "grad_norm": 5.279141194015146,
+      "learning_rate": 9.960118857070173e-06,
+      "loss": 0.21983705,
+      "memory(GiB)": 33.01,
+      "step": 5325,
+      "train_speed(iter/s)": 0.189195
+    },
+    {
+      "acc": 0.96453123,
+      "epoch": 1.7660702451954937,
+      "grad_norm": 3.180350743886554,
+      "learning_rate": 9.959946031038953e-06,
+      "loss": 0.16158001,
+      "memory(GiB)": 33.01,
+      "step": 5330,
+      "train_speed(iter/s)": 0.1892
+    },
+    {
+      "acc": 0.95384598,
+      "epoch": 1.7677269715043074,
+      "grad_norm": 4.137272848249057,
+      "learning_rate": 9.959772832849829e-06,
+      "loss": 0.19476873,
+      "memory(GiB)": 33.01,
+      "step": 5335,
+      "train_speed(iter/s)": 0.189204
+    },
+    {
+      "acc": 0.96212111,
+      "epoch": 1.7693836978131214,
+      "grad_norm": 3.018326915407665,
+      "learning_rate": 9.959599262515796e-06,
+      "loss": 0.17361283,
+      "memory(GiB)": 33.01,
+      "step": 5340,
+      "train_speed(iter/s)": 0.189208
+    },
+    {
+      "acc": 0.95687704,
+      "epoch": 1.771040424121935,
+      "grad_norm": 4.724545839141648,
+      "learning_rate": 9.959425320049883e-06,
+      "loss": 0.18612691,
+      "memory(GiB)": 33.01,
+      "step": 5345,
+      "train_speed(iter/s)": 0.189213
+    },
+    {
+      "acc": 0.95985813,
+      "epoch": 1.7726971504307487,
+      "grad_norm": 3.1952157664898784,
+      "learning_rate": 9.959251005465139e-06,
+      "loss": 0.17791543,
+      "memory(GiB)": 33.01,
+      "step": 5350,
+      "train_speed(iter/s)": 0.189217
+    },
+    {
+      "acc": 0.93917456,
+      "epoch": 1.7743538767395626,
+      "grad_norm": 7.908144443796971,
+      "learning_rate": 9.959076318774646e-06,
+      "loss": 0.29952276,
+      "memory(GiB)": 33.01,
+      "step": 5355,
+      "train_speed(iter/s)": 0.189222
+    },
+    {
+      "acc": 0.95899,
+      "epoch": 1.7760106030483764,
+      "grad_norm": 3.259262022956307,
+      "learning_rate": 9.958901259991513e-06,
+      "loss": 0.18029002,
+      "memory(GiB)": 33.01,
+      "step": 5360,
+      "train_speed(iter/s)": 0.189224
+    },
+    {
+      "acc": 0.95815773,
+      "epoch": 1.77766732935719,
+      "grad_norm": 2.3099222432152784,
+      "learning_rate": 9.958725829128877e-06,
+      "loss": 0.19149221,
+      "memory(GiB)": 33.01,
+      "step": 5365,
+      "train_speed(iter/s)": 0.189222
+    },
+    {
+      "acc": 0.96355696,
+      "epoch": 1.779324055666004,
+      "grad_norm": 3.452983714795854,
+      "learning_rate": 9.958550026199898e-06,
+      "loss": 0.19870839,
+      "memory(GiB)": 33.01,
+      "step": 5370,
+      "train_speed(iter/s)": 0.189221
+    },
+    {
+      "acc": 0.96189232,
+      "epoch": 1.7809807819748178,
+      "grad_norm": 1.6733743511786074,
+      "learning_rate": 9.958373851217772e-06,
+      "loss": 0.17930295,
+      "memory(GiB)": 33.01,
+      "step": 5375,
+      "train_speed(iter/s)": 0.189225
+    },
+    {
+      "acc": 0.97020321,
+      "epoch": 1.7826375082836314,
+      "grad_norm": 2.539753991816457,
+      "learning_rate": 9.95819730419572e-06,
+      "loss": 0.15973989,
+      "memory(GiB)": 33.01,
+      "step": 5380,
+      "train_speed(iter/s)": 0.18923
+    },
+    {
+      "acc": 0.96368361,
+      "epoch": 1.7842942345924453,
+      "grad_norm": 2.0468467926934912,
+      "learning_rate": 9.958020385146989e-06,
+      "loss": 0.19860497,
+      "memory(GiB)": 33.01,
+      "step": 5385,
+      "train_speed(iter/s)": 0.189234
+    },
+    {
+      "acc": 0.9708353,
+      "epoch": 1.7859509609012592,
+      "grad_norm": 2.8794638703337703,
+      "learning_rate": 9.957843094084853e-06,
+      "loss": 0.13811607,
+      "memory(GiB)": 33.01,
+      "step": 5390,
+      "train_speed(iter/s)": 0.189239
+    },
+    {
+      "acc": 0.96260929,
+      "epoch": 1.7876076872100728,
+      "grad_norm": 4.636465001333108,
+      "learning_rate": 9.95766543102262e-06,
+      "loss": 0.1859864,
+      "memory(GiB)": 33.01,
+      "step": 5395,
+      "train_speed(iter/s)": 0.189244
+    },
+    {
+      "acc": 0.95196228,
+      "epoch": 1.7892644135188867,
+      "grad_norm": 4.2647341048758936,
+      "learning_rate": 9.957487395973617e-06,
+      "loss": 0.21216271,
+      "memory(GiB)": 33.01,
+      "step": 5400,
+      "train_speed(iter/s)": 0.189248
+    },
+    {
+      "acc": 0.95609989,
+      "epoch": 1.7909211398277005,
+      "grad_norm": 4.495018159396596,
+      "learning_rate": 9.957308988951209e-06,
+      "loss": 0.21811047,
+      "memory(GiB)": 33.01,
+      "step": 5405,
+      "train_speed(iter/s)": 0.189253
+    },
+    {
+      "acc": 0.95574856,
+      "epoch": 1.7925778661365142,
+      "grad_norm": 3.054531740193686,
+      "learning_rate": 9.95713020996878e-06,
+      "loss": 0.20441015,
+      "memory(GiB)": 33.01,
+      "step": 5410,
+      "train_speed(iter/s)": 0.189257
+    },
+    {
+      "acc": 0.96194849,
+      "epoch": 1.794234592445328,
+      "grad_norm": 2.662537374343331,
+      "learning_rate": 9.956951059039747e-06,
+      "loss": 0.18904738,
+      "memory(GiB)": 33.01,
+      "step": 5415,
+      "train_speed(iter/s)": 0.189262
+    },
+    {
+      "acc": 0.96451588,
+      "epoch": 1.7958913187541419,
+      "grad_norm": 3.572118492814709,
+      "learning_rate": 9.956771536177553e-06,
+      "loss": 0.19641786,
+      "memory(GiB)": 33.01,
+      "step": 5420,
+      "train_speed(iter/s)": 0.189267
+    },
+    {
+      "acc": 0.96725283,
+      "epoch": 1.7975480450629555,
+      "grad_norm": 2.5921798885204437,
+      "learning_rate": 9.95659164139567e-06,
+      "loss": 0.16796941,
+      "memory(GiB)": 33.01,
+      "step": 5425,
+      "train_speed(iter/s)": 0.189272
+    },
+    {
+      "acc": 0.96207514,
+      "epoch": 1.7992047713717694,
+      "grad_norm": 3.7139254477161074,
+      "learning_rate": 9.956411374707596e-06,
+      "loss": 0.24679503,
+      "memory(GiB)": 33.01,
+      "step": 5430,
+      "train_speed(iter/s)": 0.189277
+    },
+    {
+      "acc": 0.95683517,
+      "epoch": 1.8008614976805832,
+      "grad_norm": 3.049983244753173,
+      "learning_rate": 9.956230736126861e-06,
+      "loss": 0.20307603,
+      "memory(GiB)": 33.01,
+      "step": 5435,
+      "train_speed(iter/s)": 0.189282
+    },
+    {
+      "acc": 0.97035847,
+      "epoch": 1.8025182239893969,
+      "grad_norm": 2.847732597823088,
+      "learning_rate": 9.956049725667017e-06,
+      "loss": 0.15548229,
+      "memory(GiB)": 33.01,
+      "step": 5440,
+      "train_speed(iter/s)": 0.189287
+    },
+    {
+      "acc": 0.96803818,
+      "epoch": 1.8041749502982107,
+      "grad_norm": 2.8413088988315227,
+      "learning_rate": 9.955868343341649e-06,
+      "loss": 0.16254368,
+      "memory(GiB)": 33.01,
+      "step": 5445,
+      "train_speed(iter/s)": 0.189291
+    },
+    {
+      "acc": 0.96036863,
+      "epoch": 1.8058316766070246,
+      "grad_norm": 4.110515646891542,
+      "learning_rate": 9.955686589164368e-06,
+      "loss": 0.19601402,
+      "memory(GiB)": 33.01,
+      "step": 5450,
+      "train_speed(iter/s)": 0.189296
+    },
+    {
+      "acc": 0.96154919,
+      "epoch": 1.8074884029158382,
+      "grad_norm": 2.6595176294751632,
+      "learning_rate": 9.955504463148812e-06,
+      "loss": 0.17074642,
+      "memory(GiB)": 33.01,
+      "step": 5455,
+      "train_speed(iter/s)": 0.189302
+    },
+    {
+      "acc": 0.96201286,
+      "epoch": 1.809145129224652,
+      "grad_norm": 3.7053677579403046,
+      "learning_rate": 9.955321965308649e-06,
+      "loss": 0.18650641,
+      "memory(GiB)": 33.01,
+      "step": 5460,
+      "train_speed(iter/s)": 0.189306
+    },
+    {
+      "acc": 0.96218853,
+      "epoch": 1.810801855533466,
+      "grad_norm": 3.859579308387411,
+      "learning_rate": 9.955139095657573e-06,
+      "loss": 0.17334511,
+      "memory(GiB)": 33.01,
+      "step": 5465,
+      "train_speed(iter/s)": 0.189311
+    },
+    {
+      "acc": 0.95393486,
+      "epoch": 1.8124585818422796,
+      "grad_norm": 5.1170078623053135,
+      "learning_rate": 9.954955854209306e-06,
+      "loss": 0.2070787,
+      "memory(GiB)": 33.01,
+      "step": 5470,
+      "train_speed(iter/s)": 0.189316
+    },
+    {
+      "acc": 0.9569067,
+      "epoch": 1.8141153081510935,
+      "grad_norm": 2.520467493843932,
+      "learning_rate": 9.954772240977597e-06,
+      "loss": 0.19718541,
+      "memory(GiB)": 33.01,
+      "step": 5475,
+      "train_speed(iter/s)": 0.189321
+    },
+    {
+      "acc": 0.96874485,
+      "epoch": 1.8157720344599073,
+      "grad_norm": 2.746935230366575,
+      "learning_rate": 9.954588255976228e-06,
+      "loss": 0.15265135,
+      "memory(GiB)": 33.01,
+      "step": 5480,
+      "train_speed(iter/s)": 0.189326
+    },
+    {
+      "acc": 0.95856409,
+      "epoch": 1.817428760768721,
+      "grad_norm": 3.689154169244318,
+      "learning_rate": 9.954403899219004e-06,
+      "loss": 0.18038337,
+      "memory(GiB)": 33.01,
+      "step": 5485,
+      "train_speed(iter/s)": 0.18933
+    },
+    {
+      "acc": 0.96436272,
+      "epoch": 1.8190854870775348,
+      "grad_norm": 3.482529267604087,
+      "learning_rate": 9.954219170719757e-06,
+      "loss": 0.17470418,
+      "memory(GiB)": 33.01,
+      "step": 5490,
+      "train_speed(iter/s)": 0.189336
+    },
+    {
+      "acc": 0.95511436,
+      "epoch": 1.8207422133863487,
+      "grad_norm": 3.5537877743981,
+      "learning_rate": 9.954034070492351e-06,
+      "loss": 0.20489969,
+      "memory(GiB)": 33.01,
+      "step": 5495,
+      "train_speed(iter/s)": 0.18934
+    },
+    {
+      "acc": 0.96482525,
+      "epoch": 1.8223989396951623,
+      "grad_norm": 2.419567994886498,
+      "learning_rate": 9.953848598550678e-06,
+      "loss": 0.18796244,
+      "memory(GiB)": 33.01,
+      "step": 5500,
+      "train_speed(iter/s)": 0.189345
+    },
+    {
+      "acc": 0.96336708,
+      "epoch": 1.8240556660039762,
+      "grad_norm": 3.3448878933897848,
+      "learning_rate": 9.953662754908652e-06,
+      "loss": 0.17106251,
+      "memory(GiB)": 33.01,
+      "step": 5505,
+      "train_speed(iter/s)": 0.18935
+    },
+    {
+      "acc": 0.96992645,
+      "epoch": 1.82571239231279,
+      "grad_norm": 2.7294900940660956,
+      "learning_rate": 9.95347653958022e-06,
+      "loss": 0.14668546,
+      "memory(GiB)": 33.01,
+      "step": 5510,
+      "train_speed(iter/s)": 0.189354
+    },
+    {
+      "acc": 0.95889301,
+      "epoch": 1.8273691186216037,
+      "grad_norm": 4.710714998865321,
+      "learning_rate": 9.953289952579356e-06,
+      "loss": 0.19427252,
+      "memory(GiB)": 33.01,
+      "step": 5515,
+      "train_speed(iter/s)": 0.189359
+    },
+    {
+      "acc": 0.95504227,
+      "epoch": 1.8290258449304175,
+      "grad_norm": 3.267686184607512,
+      "learning_rate": 9.953102993920061e-06,
+      "loss": 0.24130476,
+      "memory(GiB)": 33.01,
+      "step": 5520,
+      "train_speed(iter/s)": 0.189364
+    },
+    {
+      "acc": 0.95331078,
+      "epoch": 1.8306825712392314,
+      "grad_norm": 4.244190863046748,
+      "learning_rate": 9.952915663616367e-06,
+      "loss": 0.20748,
+      "memory(GiB)": 33.01,
+      "step": 5525,
+      "train_speed(iter/s)": 0.189369
+    },
+    {
+      "acc": 0.959585,
+      "epoch": 1.832339297548045,
+      "grad_norm": 5.624681869548543,
+      "learning_rate": 9.952727961682329e-06,
+      "loss": 0.21151447,
+      "memory(GiB)": 33.01,
+      "step": 5530,
+      "train_speed(iter/s)": 0.189374
+    },
+    {
+      "acc": 0.96110287,
+      "epoch": 1.833996023856859,
+      "grad_norm": 2.8042205318305466,
+      "learning_rate": 9.952539888132031e-06,
+      "loss": 0.18032423,
+      "memory(GiB)": 33.01,
+      "step": 5535,
+      "train_speed(iter/s)": 0.189378
+    },
+    {
+      "acc": 0.95601006,
+      "epoch": 1.8356527501656728,
+      "grad_norm": 3.27553391876023,
+      "learning_rate": 9.952351442979589e-06,
+      "loss": 0.22551713,
+      "memory(GiB)": 33.01,
+      "step": 5540,
+      "train_speed(iter/s)": 0.189383
+    },
+    {
+      "acc": 0.9670435,
+      "epoch": 1.8373094764744864,
+      "grad_norm": 3.4677278294009786,
+      "learning_rate": 9.95216262623914e-06,
+      "loss": 0.16397821,
+      "memory(GiB)": 33.01,
+      "step": 5545,
+      "train_speed(iter/s)": 0.189388
+    },
+    {
+      "acc": 0.9643259,
+      "epoch": 1.8389662027833003,
+      "grad_norm": 4.0715141841573566,
+      "learning_rate": 9.951973437924856e-06,
+      "loss": 0.18946757,
+      "memory(GiB)": 33.01,
+      "step": 5550,
+      "train_speed(iter/s)": 0.189393
+    },
+    {
+      "acc": 0.96782475,
+      "epoch": 1.8406229290921141,
+      "grad_norm": 3.620108181819642,
+      "learning_rate": 9.951783878050937e-06,
+      "loss": 0.15786865,
+      "memory(GiB)": 33.01,
+      "step": 5555,
+      "train_speed(iter/s)": 0.189397
+    },
+    {
+      "acc": 0.95842323,
+      "epoch": 1.8422796554009278,
+      "grad_norm": 4.018583108986213,
+      "learning_rate": 9.9515939466316e-06,
+      "loss": 0.17618065,
+      "memory(GiB)": 33.01,
+      "step": 5560,
+      "train_speed(iter/s)": 0.189402
+    },
+    {
+      "acc": 0.9770628,
+      "epoch": 1.8439363817097414,
+      "grad_norm": 3.4201700735753557,
+      "learning_rate": 9.9514036436811e-06,
+      "loss": 0.12254279,
+      "memory(GiB)": 33.01,
+      "step": 5565,
+      "train_speed(iter/s)": 0.189406
+    },
+    {
+      "acc": 0.97467213,
+      "epoch": 1.8455931080185555,
+      "grad_norm": 2.857966642793141,
+      "learning_rate": 9.951212969213721e-06,
+      "loss": 0.1302218,
+      "memory(GiB)": 33.01,
+      "step": 5570,
+      "train_speed(iter/s)": 0.189411
+    },
+    {
+      "acc": 0.97021542,
+      "epoch": 1.8472498343273691,
+      "grad_norm": 2.6506723907405125,
+      "learning_rate": 9.951021923243769e-06,
+      "loss": 0.16516323,
+      "memory(GiB)": 33.01,
+      "step": 5575,
+      "train_speed(iter/s)": 0.189415
+    },
+    {
+      "acc": 0.97295952,
+      "epoch": 1.8489065606361827,
+      "grad_norm": 4.35004800998398,
+      "learning_rate": 9.950830505785579e-06,
+      "loss": 0.16346823,
+      "memory(GiB)": 33.01,
+      "step": 5580,
+      "train_speed(iter/s)": 0.18942
+    },
+    {
+      "acc": 0.95436478,
+      "epoch": 1.8505632869449968,
+      "grad_norm": 5.718495853317044,
+      "learning_rate": 9.950638716853514e-06,
+      "loss": 0.19860299,
+      "memory(GiB)": 33.01,
+      "step": 5585,
+      "train_speed(iter/s)": 0.189422
+    },
+    {
+      "acc": 0.96391888,
+      "epoch": 1.8522200132538105,
+      "grad_norm": 2.732616807244597,
+      "learning_rate": 9.95044655646197e-06,
+      "loss": 0.16782058,
+      "memory(GiB)": 33.01,
+      "step": 5590,
+      "train_speed(iter/s)": 0.189425
+    },
+    {
+      "acc": 0.96666355,
+      "epoch": 1.853876739562624,
+      "grad_norm": 2.631055175162619,
+      "learning_rate": 9.950254024625365e-06,
+      "loss": 0.16699011,
+      "memory(GiB)": 33.01,
+      "step": 5595,
+      "train_speed(iter/s)": 0.189425
+    },
+    {
+      "acc": 0.9647212,
+      "epoch": 1.8555334658714382,
+      "grad_norm": 3.4960593654725756,
+      "learning_rate": 9.950061121358146e-06,
+      "loss": 0.17416546,
+      "memory(GiB)": 33.01,
+      "step": 5600,
+      "train_speed(iter/s)": 0.189426
+    },
+    {
+      "acc": 0.9627635,
+      "epoch": 1.8571901921802518,
+      "grad_norm": 3.4137551905665084,
+      "learning_rate": 9.949867846674787e-06,
+      "loss": 0.19350805,
+      "memory(GiB)": 33.01,
+      "step": 5605,
+      "train_speed(iter/s)": 0.18943
+    },
+    {
+      "acc": 0.96785946,
+      "epoch": 1.8588469184890655,
+      "grad_norm": 4.60939902857875,
+      "learning_rate": 9.949674200589795e-06,
+      "loss": 0.17039739,
+      "memory(GiB)": 33.01,
+      "step": 5610,
+      "train_speed(iter/s)": 0.189435
+    },
+    {
+      "acc": 0.96497545,
+      "epoch": 1.8605036447978796,
+      "grad_norm": 3.0991418877067023,
+      "learning_rate": 9.949480183117699e-06,
+      "loss": 0.16247368,
+      "memory(GiB)": 33.01,
+      "step": 5615,
+      "train_speed(iter/s)": 0.189439
+    },
+    {
+      "acc": 0.96137867,
+      "epoch": 1.8621603711066932,
+      "grad_norm": 3.78423727199372,
+      "learning_rate": 9.94928579427306e-06,
+      "loss": 0.17658565,
+      "memory(GiB)": 33.01,
+      "step": 5620,
+      "train_speed(iter/s)": 0.189444
+    },
+    {
+      "acc": 0.96758881,
+      "epoch": 1.8638170974155068,
+      "grad_norm": 2.786274762879028,
+      "learning_rate": 9.94909103407046e-06,
+      "loss": 0.1596454,
+      "memory(GiB)": 33.01,
+      "step": 5625,
+      "train_speed(iter/s)": 0.189448
+    },
+    {
+      "acc": 0.9570466,
+      "epoch": 1.8654738237243207,
+      "grad_norm": 2.4061052241565153,
+      "learning_rate": 9.948895902524519e-06,
+      "loss": 0.1833902,
+      "memory(GiB)": 33.01,
+      "step": 5630,
+      "train_speed(iter/s)": 0.189453
+    },
+    {
+      "acc": 0.9596221,
+      "epoch": 1.8671305500331346,
+      "grad_norm": 3.9073739215875074,
+      "learning_rate": 9.948700399649877e-06,
+      "loss": 0.18362517,
+      "memory(GiB)": 33.01,
+      "step": 5635,
+      "train_speed(iter/s)": 0.189457
+    },
+    {
+      "acc": 0.96014395,
+      "epoch": 1.8687872763419482,
+      "grad_norm": 4.525765170518365,
+      "learning_rate": 9.948504525461206e-06,
+      "loss": 0.18924073,
+      "memory(GiB)": 33.01,
+      "step": 5640,
+      "train_speed(iter/s)": 0.189462
+    },
+    {
+      "acc": 0.96675854,
+      "epoch": 1.870444002650762,
+      "grad_norm": 2.304845456733717,
+      "learning_rate": 9.948308279973206e-06,
+      "loss": 0.15461531,
+      "memory(GiB)": 33.01,
+      "step": 5645,
+      "train_speed(iter/s)": 0.189466
+    },
+    {
+      "acc": 0.96799221,
+      "epoch": 1.872100728959576,
+      "grad_norm": 3.1668688892631622,
+      "learning_rate": 9.9481116632006e-06,
+      "loss": 0.15957694,
+      "memory(GiB)": 33.01,
+      "step": 5650,
+      "train_speed(iter/s)": 0.189471
+    },
+    {
+      "acc": 0.96451693,
+      "epoch": 1.8737574552683895,
+      "grad_norm": 3.9604493761308337,
+      "learning_rate": 9.947914675158143e-06,
+      "loss": 0.19846363,
+      "memory(GiB)": 33.01,
+      "step": 5655,
+      "train_speed(iter/s)": 0.189475
+    },
+    {
+      "acc": 0.97062197,
+      "epoch": 1.8754141815772034,
+      "grad_norm": 4.31774606534621,
+      "learning_rate": 9.947717315860617e-06,
+      "loss": 0.15030167,
+      "memory(GiB)": 33.01,
+      "step": 5660,
+      "train_speed(iter/s)": 0.18948
+    },
+    {
+      "acc": 0.96638985,
+      "epoch": 1.8770709078860173,
+      "grad_norm": 4.0686747933870615,
+      "learning_rate": 9.947519585322832e-06,
+      "loss": 0.15593964,
+      "memory(GiB)": 33.01,
+      "step": 5665,
+      "train_speed(iter/s)": 0.189484
+    },
+    {
+      "acc": 0.96437912,
+      "epoch": 1.878727634194831,
+      "grad_norm": 3.9042899926660724,
+      "learning_rate": 9.947321483559628e-06,
+      "loss": 0.17043731,
+      "memory(GiB)": 33.01,
+      "step": 5670,
+      "train_speed(iter/s)": 0.189489
+    },
+    {
+      "acc": 0.95107632,
+      "epoch": 1.8803843605036448,
+      "grad_norm": 4.4675725921749825,
+      "learning_rate": 9.947123010585868e-06,
+      "loss": 0.21032701,
+      "memory(GiB)": 33.01,
+      "step": 5675,
+      "train_speed(iter/s)": 0.189493
+    },
+    {
+      "acc": 0.96552181,
+      "epoch": 1.8820410868124586,
+      "grad_norm": 2.2279364504453696,
+      "learning_rate": 9.946924166416445e-06,
+      "loss": 0.17136279,
+      "memory(GiB)": 33.01,
+      "step": 5680,
+      "train_speed(iter/s)": 0.189497
+    },
+    {
+      "acc": 0.96349983,
+      "epoch": 1.8836978131212723,
+      "grad_norm": 3.4830285073515905,
+      "learning_rate": 9.946724951066282e-06,
+      "loss": 0.19150252,
+      "memory(GiB)": 33.01,
+      "step": 5685,
+      "train_speed(iter/s)": 0.189502
+    },
+    {
+      "acc": 0.96433315,
+      "epoch": 1.8853545394300861,
+      "grad_norm": 2.812956752967826,
+      "learning_rate": 9.94652536455033e-06,
+      "loss": 0.17125444,
+      "memory(GiB)": 33.01,
+      "step": 5690,
+      "train_speed(iter/s)": 0.189506
+    },
+    {
+      "acc": 0.96799431,
+      "epoch": 1.8870112657389,
+      "grad_norm": 3.6819692867184632,
+      "learning_rate": 9.94632540688356e-06,
+      "loss": 0.17901212,
+      "memory(GiB)": 33.01,
+      "step": 5695,
+      "train_speed(iter/s)": 0.18951
+    },
+    {
+      "acc": 0.96423416,
+      "epoch": 1.8886679920477136,
+      "grad_norm": 3.2011576709080645,
+      "learning_rate": 9.946125078080982e-06,
+      "loss": 0.17200465,
+      "memory(GiB)": 33.01,
+      "step": 5700,
+      "train_speed(iter/s)": 0.189515
+    },
+    {
+      "acc": 0.96397667,
+      "epoch": 1.8903247183565275,
+      "grad_norm": 3.58136606188859,
+      "learning_rate": 9.945924378157627e-06,
+      "loss": 0.17308494,
+      "memory(GiB)": 33.01,
+      "step": 5705,
+      "train_speed(iter/s)": 0.189519
+    },
+    {
+      "acc": 0.97142429,
+      "epoch": 1.8919814446653413,
+      "grad_norm": 4.125904550145324,
+      "learning_rate": 9.945723307128558e-06,
+      "loss": 0.15433273,
+      "memory(GiB)": 33.01,
+      "step": 5710,
+      "train_speed(iter/s)": 0.189524
+    },
+    {
+      "acc": 0.96425629,
+      "epoch": 1.893638170974155,
+      "grad_norm": 3.413357208908564,
+      "learning_rate": 9.945521865008859e-06,
+      "loss": 0.16915461,
+      "memory(GiB)": 33.01,
+      "step": 5715,
+      "train_speed(iter/s)": 0.189528
+    },
+    {
+      "acc": 0.968606,
+      "epoch": 1.8952948972829688,
+      "grad_norm": 3.596202266452188,
+      "learning_rate": 9.94532005181365e-06,
+      "loss": 0.15530485,
+      "memory(GiB)": 33.01,
+      "step": 5720,
+      "train_speed(iter/s)": 0.189532
+    },
+    {
+      "acc": 0.96648588,
+      "epoch": 1.8969516235917827,
+      "grad_norm": 4.748709323979294,
+      "learning_rate": 9.945117867558072e-06,
+      "loss": 0.15111293,
+      "memory(GiB)": 33.01,
+      "step": 5725,
+      "train_speed(iter/s)": 0.189537
+    },
+    {
+      "acc": 0.96649303,
+      "epoch": 1.8986083499005963,
+      "grad_norm": 5.168717551643759,
+      "learning_rate": 9.9449153122573e-06,
+      "loss": 0.17624353,
+      "memory(GiB)": 33.01,
+      "step": 5730,
+      "train_speed(iter/s)": 0.189541
+    },
+    {
+      "acc": 0.9574091,
+      "epoch": 1.9002650762094102,
+      "grad_norm": 6.790787925776887,
+      "learning_rate": 9.94471238592653e-06,
+      "loss": 0.19798819,
+      "memory(GiB)": 33.01,
+      "step": 5735,
+      "train_speed(iter/s)": 0.189546
+    },
+    {
+      "acc": 0.96924324,
+      "epoch": 1.901921802518224,
+      "grad_norm": 3.632939503084288,
+      "learning_rate": 9.944509088580995e-06,
+      "loss": 0.13913025,
+      "memory(GiB)": 33.01,
+      "step": 5740,
+      "train_speed(iter/s)": 0.18955
+    },
+    {
+      "acc": 0.96660843,
+      "epoch": 1.9035785288270377,
+      "grad_norm": 2.941111655968318,
+      "learning_rate": 9.944305420235947e-06,
+      "loss": 0.16818404,
+      "memory(GiB)": 33.01,
+      "step": 5745,
+      "train_speed(iter/s)": 0.189555
+    },
+    {
+      "acc": 0.97014704,
+      "epoch": 1.9052352551358516,
+      "grad_norm": 3.759779853082322,
+      "learning_rate": 9.944101380906667e-06,
+      "loss": 0.1460492,
+      "memory(GiB)": 33.01,
+      "step": 5750,
+      "train_speed(iter/s)": 0.189559
+    },
+    {
+      "acc": 0.96233149,
+      "epoch": 1.9068919814446654,
+      "grad_norm": 2.9026664681000316,
+      "learning_rate": 9.943896970608472e-06,
+      "loss": 0.1657903,
+      "memory(GiB)": 33.01,
+      "step": 5755,
+      "train_speed(iter/s)": 0.189563
+    },
+    {
+      "acc": 0.96337318,
+      "epoch": 1.908548707753479,
+      "grad_norm": 2.7982066350813044,
+      "learning_rate": 9.943692189356696e-06,
+      "loss": 0.15194042,
+      "memory(GiB)": 33.01,
+      "step": 5760,
+      "train_speed(iter/s)": 0.189567
+    },
+    {
+      "acc": 0.96945772,
+      "epoch": 1.910205434062293,
+      "grad_norm": 3.8936747600975052,
+      "learning_rate": 9.943487037166708e-06,
+      "loss": 0.13850453,
+      "memory(GiB)": 33.01,
+      "step": 5765,
+      "train_speed(iter/s)": 0.189571
+    },
+    {
+      "acc": 0.95843439,
+      "epoch": 1.9118621603711068,
+      "grad_norm": 4.310320822323053,
+      "learning_rate": 9.943281514053902e-06,
+      "loss": 0.18423518,
+      "memory(GiB)": 33.01,
+      "step": 5770,
+      "train_speed(iter/s)": 0.189575
+    },
+    {
+      "acc": 0.96850796,
+      "epoch": 1.9135188866799204,
+      "grad_norm": 3.6817320343191,
+      "learning_rate": 9.943075620033702e-06,
+      "loss": 0.15167605,
+      "memory(GiB)": 33.01,
+      "step": 5775,
+      "train_speed(iter/s)": 0.18958
+    },
+    {
+      "acc": 0.9682394,
+      "epoch": 1.9151756129887343,
+      "grad_norm": 2.986590310742142,
+      "learning_rate": 9.942869355121556e-06,
+      "loss": 0.15726867,
+      "memory(GiB)": 33.01,
+      "step": 5780,
+      "train_speed(iter/s)": 0.189584
+    },
+    {
+      "acc": 0.96426868,
+      "epoch": 1.9168323392975481,
+      "grad_norm": 4.582984323255717,
+      "learning_rate": 9.942662719332945e-06,
+      "loss": 0.17554381,
+      "memory(GiB)": 33.01,
+      "step": 5785,
+      "train_speed(iter/s)": 0.189589
+    },
+    {
+      "acc": 0.96047783,
+      "epoch": 1.9184890656063618,
+      "grad_norm": 3.248707840523796,
+      "learning_rate": 9.942455712683373e-06,
+      "loss": 0.19423541,
+      "memory(GiB)": 33.01,
+      "step": 5790,
+      "train_speed(iter/s)": 0.189593
+    },
+    {
+      "acc": 0.96772156,
+      "epoch": 1.9201457919151756,
+      "grad_norm": 5.082307173847315,
+      "learning_rate": 9.942248335188373e-06,
+      "loss": 0.15759501,
+      "memory(GiB)": 33.01,
+      "step": 5795,
+      "train_speed(iter/s)": 0.189597
+    },
+    {
+      "acc": 0.965481,
+      "epoch": 1.9218025182239895,
+      "grad_norm": 3.4795951895953077,
+      "learning_rate": 9.94204058686351e-06,
+      "loss": 0.16091878,
+      "memory(GiB)": 33.01,
+      "step": 5800,
+      "train_speed(iter/s)": 0.189601
+    },
+    {
+      "acc": 0.96041965,
+      "epoch": 1.9234592445328031,
+      "grad_norm": 3.2883884783871697,
+      "learning_rate": 9.94183246772437e-06,
+      "loss": 0.18903304,
+      "memory(GiB)": 33.01,
+      "step": 5805,
+      "train_speed(iter/s)": 0.189606
+    },
+    {
+      "acc": 0.95328217,
+      "epoch": 1.925115970841617,
+      "grad_norm": 2.7988489264641285,
+      "learning_rate": 9.941623977786572e-06,
+      "loss": 0.21281953,
+      "memory(GiB)": 33.01,
+      "step": 5810,
+      "train_speed(iter/s)": 0.18961
+    },
+    {
+      "acc": 0.95457306,
+      "epoch": 1.9267726971504309,
+      "grad_norm": 3.389266641302172,
+      "learning_rate": 9.941415117065761e-06,
+      "loss": 0.17420642,
+      "memory(GiB)": 33.01,
+      "step": 5815,
+      "train_speed(iter/s)": 0.18961
+    },
+    {
+      "acc": 0.96181059,
+      "epoch": 1.9284294234592445,
+      "grad_norm": 4.1386874724940785,
+      "learning_rate": 9.941205885577608e-06,
+      "loss": 0.21812885,
+      "memory(GiB)": 33.01,
+      "step": 5820,
+      "train_speed(iter/s)": 0.189611
+    },
+    {
+      "acc": 0.96660843,
+      "epoch": 1.9300861497680584,
+      "grad_norm": 2.84832216281939,
+      "learning_rate": 9.940996283337817e-06,
+      "loss": 0.16445436,
+      "memory(GiB)": 33.01,
+      "step": 5825,
+      "train_speed(iter/s)": 0.18961
+    },
+    {
+      "acc": 0.96643486,
+      "epoch": 1.9317428760768722,
+      "grad_norm": 4.206598523108837,
+      "learning_rate": 9.940786310362114e-06,
+      "loss": 0.16456815,
+      "memory(GiB)": 33.01,
+      "step": 5830,
+      "train_speed(iter/s)": 0.189614
+    },
+    {
+      "acc": 0.96946182,
+      "epoch": 1.9333996023856859,
+      "grad_norm": 4.69027460844793,
+      "learning_rate": 9.940575966666259e-06,
+      "loss": 0.15309441,
+      "memory(GiB)": 33.01,
+      "step": 5835,
+      "train_speed(iter/s)": 0.189619
+    },
+    {
+      "acc": 0.97259493,
+      "epoch": 1.9350563286944995,
+      "grad_norm": 4.395042611092975,
+      "learning_rate": 9.940365252266032e-06,
+      "loss": 0.14127417,
+      "memory(GiB)": 33.01,
+      "step": 5840,
+      "train_speed(iter/s)": 0.189623
+    },
+    {
+      "acc": 0.96640015,
+      "epoch": 1.9367130550033136,
+      "grad_norm": 4.317823436957311,
+      "learning_rate": 9.940154167177245e-06,
+      "loss": 0.16580306,
+      "memory(GiB)": 33.01,
+      "step": 5845,
+      "train_speed(iter/s)": 0.189627
+    },
+    {
+      "acc": 0.95902576,
+      "epoch": 1.9383697813121272,
+      "grad_norm": 4.2769826619736735,
+      "learning_rate": 9.939942711415745e-06,
+      "loss": 0.18908883,
+      "memory(GiB)": 33.01,
+      "step": 5850,
+      "train_speed(iter/s)": 0.189631
+    },
+    {
+      "acc": 0.9634017,
+      "epoch": 1.9400265076209409,
+      "grad_norm": 3.0482091477460203,
+      "learning_rate": 9.93973088499739e-06,
+      "loss": 0.16514906,
+      "memory(GiB)": 33.01,
+      "step": 5855,
+      "train_speed(iter/s)": 0.189636
+    },
+    {
+      "acc": 0.95178919,
+      "epoch": 1.941683233929755,
+      "grad_norm": 3.310898769159179,
+      "learning_rate": 9.93951868793808e-06,
+      "loss": 0.21605191,
+      "memory(GiB)": 33.01,
+      "step": 5860,
+      "train_speed(iter/s)": 0.18964
+    },
+    {
+      "acc": 0.95677967,
+      "epoch": 1.9433399602385686,
+      "grad_norm": 2.0142825940634896,
+      "learning_rate": 9.93930612025374e-06,
+      "loss": 0.17813394,
+      "memory(GiB)": 33.01,
+      "step": 5865,
+      "train_speed(iter/s)": 0.189644
+    },
+    {
+      "acc": 0.96564751,
+      "epoch": 1.9449966865473822,
+      "grad_norm": 3.2560139068325338,
+      "learning_rate": 9.93909318196032e-06,
+      "loss": 0.16362079,
+      "memory(GiB)": 33.01,
+      "step": 5870,
+      "train_speed(iter/s)": 0.189648
+    },
+    {
+      "acc": 0.96476917,
+      "epoch": 1.9466534128561963,
+      "grad_norm": 2.8173555239865538,
+      "learning_rate": 9.938879873073794e-06,
+      "loss": 0.17951397,
+      "memory(GiB)": 33.01,
+      "step": 5875,
+      "train_speed(iter/s)": 0.189652
+    },
+    {
+      "acc": 0.96589565,
+      "epoch": 1.94831013916501,
+      "grad_norm": 3.174051439704815,
+      "learning_rate": 9.938666193610177e-06,
+      "loss": 0.14410775,
+      "memory(GiB)": 33.01,
+      "step": 5880,
+      "train_speed(iter/s)": 0.189657
+    },
+    {
+      "acc": 0.9731617,
+      "epoch": 1.9499668654738236,
+      "grad_norm": 3.8486874991213114,
+      "learning_rate": 9.938452143585498e-06,
+      "loss": 0.12638721,
+      "memory(GiB)": 33.01,
+      "step": 5885,
+      "train_speed(iter/s)": 0.189661
+    },
+    {
+      "acc": 0.97085266,
+      "epoch": 1.9516235917826377,
+      "grad_norm": 5.114970105369668,
+      "learning_rate": 9.938237723015821e-06,
+      "loss": 0.15634813,
+      "memory(GiB)": 33.01,
+      "step": 5890,
+      "train_speed(iter/s)": 0.189665
+    },
+    {
+      "acc": 0.97674122,
+      "epoch": 1.9532803180914513,
+      "grad_norm": 11.655025180971151,
+      "learning_rate": 9.938022931917237e-06,
+      "loss": 0.13883998,
+      "memory(GiB)": 33.01,
+      "step": 5895,
+      "train_speed(iter/s)": 0.189669
+    },
+    {
+      "acc": 0.96295128,
+      "epoch": 1.954937044400265,
+      "grad_norm": 4.368004065579172,
+      "learning_rate": 9.937807770305861e-06,
+      "loss": 0.18861229,
+      "memory(GiB)": 33.01,
+      "step": 5900,
+      "train_speed(iter/s)": 0.189674
+    },
+    {
+      "acc": 0.96569958,
+      "epoch": 1.9565937707090788,
+      "grad_norm": 3.3910342904685313,
+      "learning_rate": 9.937592238197843e-06,
+      "loss": 0.16494408,
+      "memory(GiB)": 33.01,
+      "step": 5905,
+      "train_speed(iter/s)": 0.189678
+    },
+    {
+      "acc": 0.95811682,
+      "epoch": 1.9582504970178927,
+      "grad_norm": 4.1514916422681045,
+      "learning_rate": 9.937376335609353e-06,
+      "loss": 0.18387959,
+      "memory(GiB)": 33.01,
+      "step": 5910,
+      "train_speed(iter/s)": 0.189682
+    },
+    {
+      "acc": 0.96629286,
+      "epoch": 1.9599072233267063,
+      "grad_norm": 2.8805711303155124,
+      "learning_rate": 9.937160062556596e-06,
+      "loss": 0.16447697,
+      "memory(GiB)": 33.01,
+      "step": 5915,
+      "train_speed(iter/s)": 0.189686
+    },
+    {
+      "acc": 0.96363659,
+      "epoch": 1.9615639496355202,
+      "grad_norm": 3.8383715650472063,
+      "learning_rate": 9.936943419055797e-06,
+      "loss": 0.16664251,
+      "memory(GiB)": 33.01,
+      "step": 5920,
+      "train_speed(iter/s)": 0.18969
+    },
+    {
+      "acc": 0.96479034,
+      "epoch": 1.963220675944334,
+      "grad_norm": 3.083885617741232,
+      "learning_rate": 9.936726405123215e-06,
+      "loss": 0.15399033,
+      "memory(GiB)": 33.01,
+      "step": 5925,
+      "train_speed(iter/s)": 0.189694
+    },
+    {
+      "acc": 0.96476812,
+      "epoch": 1.9648774022531477,
+      "grad_norm": 5.317195827197609,
+      "learning_rate": 9.936509020775135e-06,
+      "loss": 0.1703403,
+      "memory(GiB)": 33.01,
+      "step": 5930,
+      "train_speed(iter/s)": 0.189698
+    },
+    {
+      "acc": 0.96486731,
+      "epoch": 1.9665341285619615,
+      "grad_norm": 5.109553051311444,
+      "learning_rate": 9.936291266027869e-06,
+      "loss": 0.15160818,
+      "memory(GiB)": 33.01,
+      "step": 5935,
+      "train_speed(iter/s)": 0.189702
+    },
+    {
+      "acc": 0.9709569,
+      "epoch": 1.9681908548707754,
+      "grad_norm": 4.351579018153298,
+      "learning_rate": 9.936073140897757e-06,
+      "loss": 0.1614222,
+      "memory(GiB)": 33.01,
+      "step": 5940,
+      "train_speed(iter/s)": 0.189706
+    },
+    {
+      "acc": 0.96952105,
+      "epoch": 1.969847581179589,
+      "grad_norm": 3.496315852302909,
+      "learning_rate": 9.935854645401169e-06,
+      "loss": 0.16233735,
+      "memory(GiB)": 33.01,
+      "step": 5945,
+      "train_speed(iter/s)": 0.18971
+    },
+    {
+      "acc": 0.96859989,
+      "epoch": 1.9715043074884029,
+      "grad_norm": 4.402218749302247,
+      "learning_rate": 9.9356357795545e-06,
+      "loss": 0.16873567,
+      "memory(GiB)": 33.01,
+      "step": 5950,
+      "train_speed(iter/s)": 0.189714
+    },
+    {
+      "acc": 0.96829967,
+      "epoch": 1.9731610337972167,
+      "grad_norm": 5.259072901465839,
+      "learning_rate": 9.935416543374172e-06,
+      "loss": 0.19206886,
+      "memory(GiB)": 33.01,
+      "step": 5955,
+      "train_speed(iter/s)": 0.189718
+    },
+    {
+      "acc": 0.98156147,
+      "epoch": 1.9748177601060304,
+      "grad_norm": 2.9739096459866654,
+      "learning_rate": 9.93519693687664e-06,
+      "loss": 0.12224423,
+      "memory(GiB)": 33.01,
+      "step": 5960,
+      "train_speed(iter/s)": 0.189722
+    },
+    {
+      "acc": 0.96269608,
+      "epoch": 1.9764744864148442,
+      "grad_norm": 6.016239236015463,
+      "learning_rate": 9.93497696007838e-06,
+      "loss": 0.19697915,
+      "memory(GiB)": 33.01,
+      "step": 5965,
+      "train_speed(iter/s)": 0.189725
+    },
+    {
+      "acc": 0.97393789,
+      "epoch": 1.978131212723658,
+      "grad_norm": 6.86830193723295,
+      "learning_rate": 9.9347566129959e-06,
+      "loss": 0.1602066,
+      "memory(GiB)": 33.01,
+      "step": 5970,
+      "train_speed(iter/s)": 0.189729
+    },
+    {
+      "acc": 0.96911526,
+      "epoch": 1.9797879390324717,
+      "grad_norm": 4.370998411656359,
+      "learning_rate": 9.934535895645735e-06,
+      "loss": 0.16885473,
+      "memory(GiB)": 33.01,
+      "step": 5975,
+      "train_speed(iter/s)": 0.189733
+    },
+    {
+      "acc": 0.96276951,
+      "epoch": 1.9814446653412856,
+      "grad_norm": 5.0378776162092365,
+      "learning_rate": 9.93431480804445e-06,
+      "loss": 0.18467482,
+      "memory(GiB)": 33.01,
+      "step": 5980,
+      "train_speed(iter/s)": 0.189737
+    },
+    {
+      "acc": 0.9665226,
+      "epoch": 1.9831013916500995,
+      "grad_norm": 5.355865138419161,
+      "learning_rate": 9.934093350208633e-06,
+      "loss": 0.17767522,
+      "memory(GiB)": 33.01,
+      "step": 5985,
+      "train_speed(iter/s)": 0.189741
+    },
+    {
+      "acc": 0.97305756,
+      "epoch": 1.984758117958913,
+      "grad_norm": 4.467060175978817,
+      "learning_rate": 9.933871522154901e-06,
+      "loss": 0.14633491,
+      "memory(GiB)": 33.01,
+      "step": 5990,
+      "train_speed(iter/s)": 0.189745
+    },
+    {
+      "acc": 0.96583843,
+      "epoch": 1.986414844267727,
+      "grad_norm": 2.93855599647747,
+      "learning_rate": 9.933649323899903e-06,
+      "loss": 0.14597803,
+      "memory(GiB)": 33.01,
+      "step": 5995,
+      "train_speed(iter/s)": 0.189749
+    },
+    {
+      "acc": 0.97019806,
+      "epoch": 1.9880715705765408,
+      "grad_norm": 5.470550197007648,
+      "learning_rate": 9.933426755460312e-06,
+      "loss": 0.16220329,
+      "memory(GiB)": 33.01,
+      "step": 6000,
+      "train_speed(iter/s)": 0.189753
+    },
+    {
+      "acc": 0.98152065,
+      "epoch": 1.9897282968853545,
+      "grad_norm": 2.4884927713641116,
+      "learning_rate": 9.933203816852828e-06,
+      "loss": 0.10297159,
+      "memory(GiB)": 33.01,
+      "step": 6005,
+      "train_speed(iter/s)": 0.189757
+    },
+    {
+      "acc": 0.96945152,
+      "epoch": 1.9913850231941683,
+      "grad_norm": 2.2605004743369124,
+      "learning_rate": 9.932980508094182e-06,
+      "loss": 0.13022065,
+      "memory(GiB)": 33.01,
+      "step": 6010,
+      "train_speed(iter/s)": 0.189761
+    },
+    {
+      "acc": 0.96832008,
+      "epoch": 1.9930417495029822,
+      "grad_norm": 4.646889690280367,
+      "learning_rate": 9.93275682920113e-06,
+      "loss": 0.16072577,
+      "memory(GiB)": 33.01,
+      "step": 6015,
+      "train_speed(iter/s)": 0.189765
+    },
+    {
+      "acc": 0.96582203,
+      "epoch": 1.9946984758117958,
+      "grad_norm": 4.175424237969419,
+      "learning_rate": 9.932532780190458e-06,
+      "loss": 0.16094139,
+      "memory(GiB)": 33.01,
+      "step": 6020,
+      "train_speed(iter/s)": 0.189769
+    },
+    {
+      "acc": 0.96637249,
+      "epoch": 1.9963552021206097,
+      "grad_norm": 4.636797532893509,
+      "learning_rate": 9.932308361078978e-06,
+      "loss": 0.16655571,
+      "memory(GiB)": 33.01,
+      "step": 6025,
+      "train_speed(iter/s)": 0.189773
+    },
+    {
+      "acc": 0.9627594,
+      "epoch": 1.9980119284294235,
+      "grad_norm": 3.059624579163151,
+      "learning_rate": 9.932083571883532e-06,
+      "loss": 0.18198231,
+      "memory(GiB)": 33.01,
+      "step": 6030,
+      "train_speed(iter/s)": 0.189777
+    },
+    {
+      "acc": 0.96720581,
+      "epoch": 1.9996686547382372,
+      "grad_norm": 3.3566397258529785,
+      "learning_rate": 9.931858412620986e-06,
+      "loss": 0.15749495,
+      "memory(GiB)": 33.01,
+      "step": 6035,
+      "train_speed(iter/s)": 0.189782
+    },
+    {
+      "epoch": 2.0,
+      "eval_acc": 0.9713093375065206,
+      "eval_loss": 0.07300830632448196,
+      "eval_runtime": 508.1599,
+      "eval_samples_per_second": 10.558,
+      "eval_steps_per_second": 1.32,
+      "step": 6036
+    },
+    {
+      "acc": 0.41311496,
+      "epoch": 10.901713255184852,
+      "grad_norm": 46.142024967492716,
+      "learning_rate": 4.667995707060734e-06,
+      "loss": 18.62856674,
+      "memory(GiB)": 34.67,
+      "step": 6040,
+      "train_speed(iter/s)": 134.249727
+    },
+    {
+      "acc": 0.46514187,
+      "epoch": 10.910730387736699,
+      "grad_norm": 24.64683148399521,
+      "learning_rate": 4.66055183850045e-06,
+      "loss": 8.34107056,
+      "memory(GiB)": 34.67,
+      "step": 6045,
+      "train_speed(iter/s)": 81.484968
+    },
+    {
+      "acc": 0.5982141,
+      "epoch": 10.919747520288547,
+      "grad_norm": 23.668328342267483,
+      "learning_rate": 4.653108726992041e-06,
+      "loss": 4.0812561,
+      "memory(GiB)": 34.67,
+      "step": 6050,
+      "train_speed(iter/s)": 58.456234
+    },
+    {
+      "acc": 0.76831255,
+      "epoch": 10.928764652840396,
+      "grad_norm": 8.75274323714371,
+      "learning_rate": 4.645666389111036e-06,
+      "loss": 1.87636223,
+      "memory(GiB)": 34.67,
+      "step": 6055,
+      "train_speed(iter/s)": 45.601259
+    },
+    {
+      "acc": 0.77068529,
+      "epoch": 10.937781785392245,
+      "grad_norm": 7.157418515237572,
+      "learning_rate": 4.638224841431247e-06,
+      "loss": 1.38030567,
+      "memory(GiB)": 34.67,
+      "step": 6060,
+      "train_speed(iter/s)": 37.348973
+    },
+    {
+      "acc": 0.82335567,
+      "epoch": 10.946798917944093,
+      "grad_norm": 6.404999379717916,
+      "learning_rate": 4.630784100524721e-06,
+      "loss": 1.05846272,
+      "memory(GiB)": 34.67,
+      "step": 6065,
+      "train_speed(iter/s)": 31.659352
+    },
+    {
+      "acc": 0.81055689,
+      "epoch": 10.955816050495942,
+      "grad_norm": 5.668317115514912,
+      "learning_rate": 4.623344182961712e-06,
+      "loss": 1.10004044,
+      "memory(GiB)": 34.67,
+      "step": 6070,
+      "train_speed(iter/s)": 27.482731
+    },
+    {
+      "acc": 0.81895657,
+      "epoch": 10.96483318304779,
+      "grad_norm": 5.1716632093404105,
+      "learning_rate": 4.615905105310636e-06,
+      "loss": 1.02108898,
+      "memory(GiB)": 34.67,
+      "step": 6075,
+      "train_speed(iter/s)": 24.283442
+    },
+    {
+      "acc": 0.82773285,
+      "epoch": 10.97385031559964,
+      "grad_norm": 8.46597160197095,
+      "learning_rate": 4.608466884138046e-06,
+      "loss": 1.06346617,
+      "memory(GiB)": 34.67,
+      "step": 6080,
+      "train_speed(iter/s)": 21.750531
+    },
+    {
+      "acc": 0.85879459,
+      "epoch": 10.982867448151488,
+      "grad_norm": 7.763148888195122,
+      "learning_rate": 4.601029536008578e-06,
+      "loss": 0.83358173,
+      "memory(GiB)": 34.67,
+      "step": 6085,
+      "train_speed(iter/s)": 19.702423
+    },
+    {
+      "acc": 0.85695562,
+      "epoch": 10.991884580703337,
+      "grad_norm": 3.3846154403296946,
+      "learning_rate": 4.593593077484933e-06,
+      "loss": 0.80661335,
+      "memory(GiB)": 34.67,
+      "step": 6090,
+      "train_speed(iter/s)": 18.009462
+    },
+    {
+      "epoch": 10.999098286744815,
+      "eval_acc": 0.8568162926018288,
+      "eval_loss": 0.4145449101924896,
+      "eval_runtime": 49.8486,
+      "eval_samples_per_second": 9.368,
+      "eval_steps_per_second": 1.184,
+      "step": 6094
+    },
+    {
+      "acc": 0.95750618,
+      "epoch": 11.001803426510369,
+      "grad_norm": 4.465321488619032,
+      "learning_rate": 4.586157525127823e-06,
+      "loss": 0.84973698,
+      "memory(GiB)": 34.67,
+      "step": 6095,
+      "train_speed(iter/s)": 12.899899
+    },
+    {
+      "acc": 0.86048241,
+      "epoch": 11.010820559062218,
+      "grad_norm": 7.565049725009297,
+      "learning_rate": 4.57872289549595e-06,
+      "loss": 0.87631731,
+      "memory(GiB)": 34.67,
+      "step": 6100,
+      "train_speed(iter/s)": 12.156184
+    },
+    {
+      "acc": 0.86794395,
+      "epoch": 11.019837691614066,
+      "grad_norm": 4.406828723959708,
+      "learning_rate": 4.5712892051459525e-06,
+      "loss": 0.77457833,
+      "memory(GiB)": 34.67,
+      "step": 6105,
+      "train_speed(iter/s)": 11.495281
+    },
+    {
+      "acc": 0.85152607,
+      "epoch": 11.028854824165915,
+      "grad_norm": 4.05923810617219,
+      "learning_rate": 4.563856470632386e-06,
+      "loss": 0.86453714,
+      "memory(GiB)": 34.67,
+      "step": 6110,
+      "train_speed(iter/s)": 10.90187
+    },
+    {
+      "acc": 0.87472191,
+      "epoch": 11.037871956717764,
+      "grad_norm": 4.997244876473967,
+      "learning_rate": 4.556424708507669e-06,
+      "loss": 0.73219786,
+      "memory(GiB)": 34.67,
+      "step": 6115,
+      "train_speed(iter/s)": 10.366957
+    },
+    {
+      "acc": 0.86167145,
+      "epoch": 11.046889089269612,
+      "grad_norm": 4.27729182639188,
+      "learning_rate": 4.548993935322059e-06,
+      "loss": 0.80507736,
+      "memory(GiB)": 34.67,
+      "step": 6120,
+      "train_speed(iter/s)": 9.883493
+    },
+    {
+      "acc": 0.84846478,
+      "epoch": 11.055906221821461,
+      "grad_norm": 3.3910721001118187,
+      "learning_rate": 4.541564167623612e-06,
+      "loss": 0.8367732,
+      "memory(GiB)": 34.67,
+      "step": 6125,
+      "train_speed(iter/s)": 9.44275
+    },
+    {
+      "acc": 0.85515461,
+      "epoch": 11.06492335437331,
+      "grad_norm": 3.7405187429618736,
+      "learning_rate": 4.534135421958145e-06,
+      "loss": 0.81323128,
+      "memory(GiB)": 34.67,
+      "step": 6130,
+      "train_speed(iter/s)": 9.040436
+    },
+    {
+      "acc": 0.85793438,
+      "epoch": 11.073940486925158,
+      "grad_norm": 3.089469841978783,
+      "learning_rate": 4.526707714869195e-06,
+      "loss": 0.7354743,
+      "memory(GiB)": 34.67,
+      "step": 6135,
+      "train_speed(iter/s)": 8.672552
+    },
+    {
+      "acc": 0.88629551,
+      "epoch": 11.082957619477007,
+      "grad_norm": 7.313152698406621,
+      "learning_rate": 4.51928106289799e-06,
+      "loss": 0.61730213,
+      "memory(GiB)": 34.67,
+      "step": 6140,
+      "train_speed(iter/s)": 8.333162
+    },
+    {
+      "acc": 0.88074913,
+      "epoch": 11.091974752028856,
+      "grad_norm": 4.177627478142328,
+      "learning_rate": 4.511855482583406e-06,
+      "loss": 0.66737604,
+      "memory(GiB)": 34.67,
+      "step": 6145,
+      "train_speed(iter/s)": 8.020784
+    },
+    {
+      "acc": 0.87245197,
+      "epoch": 11.100991884580703,
+      "grad_norm": 8.500306095037253,
+      "learning_rate": 4.504430990461934e-06,
+      "loss": 0.73731365,
+      "memory(GiB)": 34.67,
+      "step": 6150,
+      "train_speed(iter/s)": 7.730995
+    },
+    {
+      "acc": 0.85932751,
+      "epoch": 11.110009017132551,
+      "grad_norm": 5.635892801921693,
+      "learning_rate": 4.49700760306764e-06,
+      "loss": 0.76138706,
+      "memory(GiB)": 34.67,
+      "step": 6155,
+      "train_speed(iter/s)": 7.46189
+    },
+    {
+      "acc": 0.86015644,
+      "epoch": 11.1190261496844,
+      "grad_norm": 5.5551702753791785,
+      "learning_rate": 4.489585336932132e-06,
+      "loss": 0.87525578,
+      "memory(GiB)": 34.67,
+      "step": 6160,
+      "train_speed(iter/s)": 7.21102
+    },
+    {
+      "acc": 0.85932522,
+      "epoch": 11.128043282236249,
+      "grad_norm": 4.989421229267309,
+      "learning_rate": 4.482164208584518e-06,
+      "loss": 0.73783965,
+      "memory(GiB)": 34.67,
+      "step": 6165,
+      "train_speed(iter/s)": 6.976535
+    },
+    {
+      "acc": 0.84846373,
+      "epoch": 11.137060414788097,
+      "grad_norm": 6.395565631558778,
+      "learning_rate": 4.474744234551378e-06,
+      "loss": 0.84804173,
+      "memory(GiB)": 34.67,
+      "step": 6170,
+      "train_speed(iter/s)": 6.757025
+    },
+    {
+      "acc": 0.86683969,
+      "epoch": 11.146077547339946,
+      "grad_norm": 4.076153166920833,
+      "learning_rate": 4.467325431356711e-06,
+      "loss": 0.7369915,
+      "memory(GiB)": 34.67,
+      "step": 6175,
+      "train_speed(iter/s)": 6.551458
+    },
+    {
+      "acc": 0.89684353,
+      "epoch": 11.155094679891794,
+      "grad_norm": 3.1160145079663875,
+      "learning_rate": 4.459907815521921e-06,
+      "loss": 0.56216383,
+      "memory(GiB)": 34.67,
+      "step": 6180,
+      "train_speed(iter/s)": 6.351374
+    },
+    {
+      "acc": 0.88666992,
+      "epoch": 11.164111812443643,
+      "grad_norm": 5.954337282924469,
+      "learning_rate": 4.452491403565758e-06,
+      "loss": 0.64935493,
+      "memory(GiB)": 34.67,
+      "step": 6185,
+      "train_speed(iter/s)": 6.167535
+    },
+    {
+      "acc": 0.89458447,
+      "epoch": 11.173128944995492,
+      "grad_norm": 6.2738775389672385,
+      "learning_rate": 4.445076212004297e-06,
+      "loss": 0.60745091,
+      "memory(GiB)": 34.67,
+      "step": 6190,
+      "train_speed(iter/s)": 5.996881
+    },
+    {
+      "acc": 0.8762886,
+      "epoch": 11.18214607754734,
+      "grad_norm": 3.591047627327035,
+      "learning_rate": 4.437662257350892e-06,
+      "loss": 0.70382156,
+      "memory(GiB)": 34.67,
+      "step": 6195,
+      "train_speed(iter/s)": 5.835641
+    },
+    {
+      "acc": 0.86773987,
+      "epoch": 11.19116321009919,
+      "grad_norm": 5.914609622702315,
+      "learning_rate": 4.430249556116143e-06,
+      "loss": 0.75445595,
+      "memory(GiB)": 34.67,
+      "step": 6200,
+      "train_speed(iter/s)": 5.682762
+    },
+    {
+      "acc": 0.86218958,
+      "epoch": 11.200180342651038,
+      "grad_norm": 4.554153845408479,
+      "learning_rate": 4.42283812480786e-06,
+      "loss": 0.80199242,
+      "memory(GiB)": 34.67,
+      "step": 6205,
+      "train_speed(iter/s)": 5.538199
+    },
+    {
+      "acc": 0.89075756,
+      "epoch": 11.209197475202885,
+      "grad_norm": 5.4323685656726175,
+      "learning_rate": 4.415427979931023e-06,
+      "loss": 0.61459293,
+      "memory(GiB)": 34.67,
+      "step": 6210,
+      "train_speed(iter/s)": 5.401332
+    },
+    {
+      "acc": 0.87350368,
+      "epoch": 11.218214607754733,
+      "grad_norm": 4.332978436405154,
+      "learning_rate": 4.40801913798775e-06,
+      "loss": 0.76399531,
+      "memory(GiB)": 34.67,
+      "step": 6215,
+      "train_speed(iter/s)": 5.270833
+    },
+    {
+      "acc": 0.8791748,
+      "epoch": 11.227231740306582,
+      "grad_norm": 3.2931449207102945,
+      "learning_rate": 4.400611615477254e-06,
+      "loss": 0.70324554,
+      "memory(GiB)": 34.67,
+      "step": 6220,
+      "train_speed(iter/s)": 5.146818
+    },
+    {
+      "acc": 0.90632401,
+      "epoch": 11.23624887285843,
+      "grad_norm": 3.786264171328009,
+      "learning_rate": 4.393205428895812e-06,
+      "loss": 0.54246979,
+      "memory(GiB)": 34.67,
+      "step": 6225,
+      "train_speed(iter/s)": 5.028875
+    },
+    {
+      "acc": 0.87950859,
+      "epoch": 11.24526600541028,
+      "grad_norm": 4.913441290267515,
+      "learning_rate": 4.385800594736724e-06,
+      "loss": 0.6850523,
+      "memory(GiB)": 34.67,
+      "step": 6230,
+      "train_speed(iter/s)": 4.916327
+    },
+    {
+      "acc": 0.89150009,
+      "epoch": 11.254283137962128,
+      "grad_norm": 6.077508029370644,
+      "learning_rate": 4.37839712949028e-06,
+      "loss": 0.63446875,
+      "memory(GiB)": 34.67,
+      "step": 6235,
+      "train_speed(iter/s)": 4.808878
+    },
+    {
+      "acc": 0.91192722,
+      "epoch": 11.263300270513977,
+      "grad_norm": 3.003061954894193,
+      "learning_rate": 4.370995049643723e-06,
+      "loss": 0.55151949,
+      "memory(GiB)": 34.67,
+      "step": 6240,
+      "train_speed(iter/s)": 4.706148
+    },
+    {
+      "acc": 0.87685108,
+      "epoch": 11.272317403065825,
+      "grad_norm": 4.869795131771389,
+      "learning_rate": 4.363594371681206e-06,
+      "loss": 0.71072888,
+      "memory(GiB)": 34.67,
+      "step": 6245,
+      "train_speed(iter/s)": 4.608017
+    },
+    {
+      "acc": 0.87757835,
+      "epoch": 11.281334535617674,
+      "grad_norm": 3.144467522698747,
+      "learning_rate": 4.356195112083764e-06,
+      "loss": 0.68076372,
+      "memory(GiB)": 34.67,
+      "step": 6250,
+      "train_speed(iter/s)": 4.513877
+    },
+    {
+      "acc": 0.87459698,
+      "epoch": 11.290351668169523,
+      "grad_norm": 3.4056717021233385,
+      "learning_rate": 4.348797287329274e-06,
+      "loss": 0.66040697,
+      "memory(GiB)": 34.67,
+      "step": 6255,
+      "train_speed(iter/s)": 4.423773
+    },
+    {
+      "acc": 0.89757004,
+      "epoch": 11.299368800721371,
+      "grad_norm": 3.0460819816788094,
+      "learning_rate": 4.341400913892411e-06,
+      "loss": 0.57552347,
+      "memory(GiB)": 34.67,
+      "step": 6260,
+      "train_speed(iter/s)": 4.337417
+    },
+    {
+      "acc": 0.87836895,
+      "epoch": 11.30838593327322,
+      "grad_norm": 3.362601582141929,
+      "learning_rate": 4.3340060082446275e-06,
+      "loss": 0.69174132,
+      "memory(GiB)": 34.67,
+      "step": 6265,
+      "train_speed(iter/s)": 4.254369
+    },
+    {
+      "acc": 0.89809933,
+      "epoch": 11.317403065825067,
+      "grad_norm": 3.5130130069687056,
+      "learning_rate": 4.326612586854099e-06,
+      "loss": 0.625875,
+      "memory(GiB)": 34.67,
+      "step": 6270,
+      "train_speed(iter/s)": 4.174463
+    },
+    {
+      "acc": 0.87767344,
+      "epoch": 11.326420198376915,
+      "grad_norm": 3.5824019523244695,
+      "learning_rate": 4.319220666185701e-06,
+      "loss": 0.68774424,
+      "memory(GiB)": 34.67,
+      "step": 6275,
+      "train_speed(iter/s)": 4.097551
+    },
+    {
+      "acc": 0.86709652,
+      "epoch": 11.335437330928764,
+      "grad_norm": 4.182412722715768,
+      "learning_rate": 4.311830262700964e-06,
+      "loss": 0.70857172,
+      "memory(GiB)": 34.67,
+      "step": 6280,
+      "train_speed(iter/s)": 4.023776
+    },
+    {
+      "acc": 0.88084459,
+      "epoch": 11.344454463480613,
+      "grad_norm": 5.101241830487778,
+      "learning_rate": 4.304441392858043e-06,
+      "loss": 0.64174843,
+      "memory(GiB)": 34.67,
+      "step": 6285,
+      "train_speed(iter/s)": 3.952641
+    },
+    {
+      "acc": 0.90726929,
+      "epoch": 11.353471596032461,
+      "grad_norm": 9.160278188310272,
+      "learning_rate": 4.2970540731116725e-06,
+      "loss": 0.51663237,
+      "memory(GiB)": 34.67,
+      "step": 6290,
+      "train_speed(iter/s)": 3.884066
+    },
+    {
+      "acc": 0.8933569,
+      "epoch": 11.36248872858431,
+      "grad_norm": 3.085642138490934,
+      "learning_rate": 4.28966831991314e-06,
+      "loss": 0.57848482,
+      "memory(GiB)": 34.67,
+      "step": 6295,
+      "train_speed(iter/s)": 3.817883
+    },
+    {
+      "acc": 0.87269354,
+      "epoch": 11.371505861136159,
+      "grad_norm": 5.681723562526792,
+      "learning_rate": 4.282284149710242e-06,
+      "loss": 0.64417229,
+      "memory(GiB)": 34.67,
+      "step": 6300,
+      "train_speed(iter/s)": 3.754185
+    },
+    {
+      "acc": 0.87589941,
+      "epoch": 11.380522993688007,
+      "grad_norm": 3.431820796043091,
+      "learning_rate": 4.27490157894725e-06,
+      "loss": 0.64798918,
+      "memory(GiB)": 34.67,
+      "step": 6305,
+      "train_speed(iter/s)": 3.692593
+    },
+    {
+      "acc": 0.89470549,
+      "epoch": 11.389540126239856,
+      "grad_norm": 3.4942441599389817,
+      "learning_rate": 4.267520624064872e-06,
+      "loss": 0.59482164,
+      "memory(GiB)": 34.67,
+      "step": 6310,
+      "train_speed(iter/s)": 3.633047
+    },
+    {
+      "acc": 0.91079884,
+      "epoch": 11.398557258791705,
+      "grad_norm": 3.3775386764524638,
+      "learning_rate": 4.260141301500224e-06,
+      "loss": 0.48047276,
+      "memory(GiB)": 34.67,
+      "step": 6315,
+      "train_speed(iter/s)": 3.575415
+    },
+    {
+      "acc": 0.88378391,
+      "epoch": 11.407574391343553,
+      "grad_norm": 4.1693750321585314,
+      "learning_rate": 4.2527636276867765e-06,
+      "loss": 0.62509356,
+      "memory(GiB)": 34.67,
+      "step": 6320,
+      "train_speed(iter/s)": 3.519721
+    },
+    {
+      "acc": 0.8680604,
+      "epoch": 11.416591523895402,
+      "grad_norm": 2.759915992870974,
+      "learning_rate": 4.245387619054336e-06,
+      "loss": 0.69436426,
+      "memory(GiB)": 34.67,
+      "step": 6325,
+      "train_speed(iter/s)": 3.465826
+    },
+    {
+      "acc": 0.86723642,
+      "epoch": 11.425608656447249,
+      "grad_norm": 4.066922897425721,
+      "learning_rate": 4.238013292028999e-06,
+      "loss": 0.6587019,
+      "memory(GiB)": 34.67,
+      "step": 6330,
+      "train_speed(iter/s)": 3.413544
+    },
+    {
+      "acc": 0.89130516,
+      "epoch": 11.434625788999098,
+      "grad_norm": 2.7832289877775747,
+      "learning_rate": 4.230640663033118e-06,
+      "loss": 0.62211418,
+      "memory(GiB)": 34.67,
+      "step": 6335,
+      "train_speed(iter/s)": 3.362958
+    },
+    {
+      "acc": 0.86635342,
+      "epoch": 11.443642921550946,
+      "grad_norm": 3.3555983359646153,
+      "learning_rate": 4.223269748485257e-06,
+      "loss": 0.72348833,
+      "memory(GiB)": 34.67,
+      "step": 6340,
+      "train_speed(iter/s)": 3.313855
+    },
+    {
+      "acc": 0.8849308,
+      "epoch": 11.452660054102795,
+      "grad_norm": 3.4129318034164076,
+      "learning_rate": 4.2159005648001764e-06,
+      "loss": 0.64281712,
+      "memory(GiB)": 34.67,
+      "step": 6345,
+      "train_speed(iter/s)": 3.266216
+    },
+    {
+      "acc": 0.88074913,
+      "epoch": 11.461677186654644,
+      "grad_norm": 3.194855967646834,
+      "learning_rate": 4.208533128388767e-06,
+      "loss": 0.62555776,
+      "memory(GiB)": 34.67,
+      "step": 6350,
+      "train_speed(iter/s)": 3.220086
+    },
+    {
+      "acc": 0.88982649,
+      "epoch": 11.470694319206492,
+      "grad_norm": 3.317925136227602,
+      "learning_rate": 4.201167455658038e-06,
+      "loss": 0.5781611,
+      "memory(GiB)": 34.67,
+      "step": 6355,
+      "train_speed(iter/s)": 3.175256
+    },
+    {
+      "acc": 0.88328705,
+      "epoch": 11.479711451758341,
+      "grad_norm": 3.483165660746629,
+      "learning_rate": 4.193803563011065e-06,
+      "loss": 0.68382282,
+      "memory(GiB)": 34.67,
+      "step": 6360,
+      "train_speed(iter/s)": 3.131704
+    },
+    {
+      "acc": 0.89734564,
+      "epoch": 11.48872858431019,
+      "grad_norm": 3.9351604916671716,
+      "learning_rate": 4.186441466846966e-06,
+      "loss": 0.53999434,
+      "memory(GiB)": 34.67,
+      "step": 6365,
+      "train_speed(iter/s)": 3.089461
+    },
+    {
+      "acc": 0.90035248,
+      "epoch": 11.497745716862038,
+      "grad_norm": 3.457828710313778,
+      "learning_rate": 4.17908118356085e-06,
+      "loss": 0.58038116,
+      "memory(GiB)": 34.67,
+      "step": 6370,
+      "train_speed(iter/s)": 3.048327
+    },
+    {
+      "acc": 0.90201874,
+      "epoch": 11.506762849413887,
+      "grad_norm": 4.529124402650824,
+      "learning_rate": 4.171722729543797e-06,
+      "loss": 0.50730953,
+      "memory(GiB)": 34.67,
+      "step": 6375,
+      "train_speed(iter/s)": 3.008356
+    },
+    {
+      "acc": 0.89611263,
+      "epoch": 11.515779981965736,
+      "grad_norm": 3.523044991804961,
+      "learning_rate": 4.164366121182807e-06,
+      "loss": 0.58241019,
+      "memory(GiB)": 34.67,
+      "step": 6380,
+      "train_speed(iter/s)": 2.969551
+    },
+    {
+      "acc": 0.90218925,
+      "epoch": 11.524797114517582,
+      "grad_norm": 2.8492217525856893,
+      "learning_rate": 4.157011374860774e-06,
+      "loss": 0.53789234,
+      "memory(GiB)": 34.67,
+      "step": 6385,
+      "train_speed(iter/s)": 2.931828
+    },
+    {
+      "acc": 0.89348812,
+      "epoch": 11.533814247069431,
+      "grad_norm": 3.2999194874628395,
+      "learning_rate": 4.149658506956444e-06,
+      "loss": 0.55231614,
+      "memory(GiB)": 34.67,
+      "step": 6390,
+      "train_speed(iter/s)": 2.895088
+    },
+    {
+      "acc": 0.88687525,
+      "epoch": 11.54283137962128,
+      "grad_norm": 5.519207580798505,
+      "learning_rate": 4.142307533844378e-06,
+      "loss": 0.5719964,
+      "memory(GiB)": 34.67,
+      "step": 6395,
+      "train_speed(iter/s)": 2.859315
+    },
+    {
+      "acc": 0.87307539,
+      "epoch": 11.551848512173128,
+      "grad_norm": 3.2697999679942833,
+      "learning_rate": 4.13495847189492e-06,
+      "loss": 0.63465419,
+      "memory(GiB)": 34.67,
+      "step": 6400,
+      "train_speed(iter/s)": 2.824461
+    },
+    {
+      "acc": 0.89108429,
+      "epoch": 11.560865644724977,
+      "grad_norm": 3.439456573595362,
+      "learning_rate": 4.1276113374741555e-06,
+      "loss": 0.60180283,
+      "memory(GiB)": 34.67,
+      "step": 6405,
+      "train_speed(iter/s)": 2.790087
+    },
+    {
+      "acc": 0.88042135,
+      "epoch": 11.569882777276826,
+      "grad_norm": 2.7209605662247496,
+      "learning_rate": 4.120266146943882e-06,
+      "loss": 0.65857315,
+      "memory(GiB)": 34.67,
+      "step": 6410,
+      "train_speed(iter/s)": 2.756114
+    },
+    {
+      "acc": 0.88512735,
+      "epoch": 11.578899909828674,
+      "grad_norm": 3.687439173536713,
+      "learning_rate": 4.112922916661561e-06,
+      "loss": 0.63806605,
+      "memory(GiB)": 34.67,
+      "step": 6415,
+      "train_speed(iter/s)": 2.723805
+    },
+    {
+      "acc": 0.89790583,
+      "epoch": 11.587917042380523,
+      "grad_norm": 5.283415110789133,
+      "learning_rate": 4.105581662980295e-06,
+      "loss": 0.5472147,
+      "memory(GiB)": 34.67,
+      "step": 6420,
+      "train_speed(iter/s)": 2.692311
+    },
+    {
+      "acc": 0.89327641,
+      "epoch": 11.596934174932372,
+      "grad_norm": 2.171938859098593,
+      "learning_rate": 4.098242402248779e-06,
+      "loss": 0.58158698,
+      "memory(GiB)": 34.67,
+      "step": 6425,
+      "train_speed(iter/s)": 2.661569
+    },
+    {
+      "acc": 0.88704071,
+      "epoch": 11.60595130748422,
+      "grad_norm": 3.773846594518904,
+      "learning_rate": 4.090905150811277e-06,
+      "loss": 0.56592655,
+      "memory(GiB)": 34.67,
+      "step": 6430,
+      "train_speed(iter/s)": 2.631583
+    },
+    {
+      "acc": 0.88291311,
+      "epoch": 11.614968440036069,
+      "grad_norm": 9.70034059919439,
+      "learning_rate": 4.083569925007571e-06,
+      "loss": 0.61799598,
+      "memory(GiB)": 34.67,
+      "step": 6435,
+      "train_speed(iter/s)": 2.602311
+    },
+    {
+      "acc": 0.8969924,
+      "epoch": 11.623985572587918,
+      "grad_norm": 2.46772646889123,
+      "learning_rate": 4.076236741172937e-06,
+      "loss": 0.57910976,
+      "memory(GiB)": 34.67,
+      "step": 6440,
+      "train_speed(iter/s)": 2.573708
+    },
+    {
+      "acc": 0.89131985,
+      "epoch": 11.633002705139766,
+      "grad_norm": 2.839504986424735,
+      "learning_rate": 4.068905615638101e-06,
+      "loss": 0.55641756,
+      "memory(GiB)": 34.67,
+      "step": 6445,
+      "train_speed(iter/s)": 2.54581
+    },
+    {
+      "acc": 0.89318829,
+      "epoch": 11.642019837691613,
+      "grad_norm": 3.738850204580844,
+      "learning_rate": 4.061576564729205e-06,
+      "loss": 0.58655119,
+      "memory(GiB)": 34.67,
+      "step": 6450,
+      "train_speed(iter/s)": 2.518519
+    },
+    {
+      "acc": 0.8902998,
+      "epoch": 11.651036970243462,
+      "grad_norm": 3.086625314999831,
+      "learning_rate": 4.0542496047677736e-06,
+      "loss": 0.56620655,
+      "memory(GiB)": 34.67,
+      "step": 6455,
+      "train_speed(iter/s)": 2.49192
+    },
+    {
+      "acc": 0.90637522,
+      "epoch": 11.66005410279531,
+      "grad_norm": 2.7876811806222728,
+      "learning_rate": 4.046924752070674e-06,
+      "loss": 0.52601414,
+      "memory(GiB)": 34.67,
+      "step": 6460,
+      "train_speed(iter/s)": 2.465954
+    },
+    {
+      "acc": 0.88894434,
+      "epoch": 11.66907123534716,
+      "grad_norm": 2.7121026127735903,
+      "learning_rate": 4.03960202295008e-06,
+      "loss": 0.58678231,
+      "memory(GiB)": 34.67,
+      "step": 6465,
+      "train_speed(iter/s)": 2.44045
+    },
+    {
+      "acc": 0.90352039,
+      "epoch": 11.678088367899008,
+      "grad_norm": 2.8581865030348097,
+      "learning_rate": 4.032281433713433e-06,
+      "loss": 0.5328721,
+      "memory(GiB)": 34.67,
+      "step": 6470,
+      "train_speed(iter/s)": 2.41557
+    },
+    {
+      "acc": 0.90520477,
+      "epoch": 11.687105500450857,
+      "grad_norm": 2.1320453957594507,
+      "learning_rate": 4.024963000663416e-06,
+      "loss": 0.53016138,
+      "memory(GiB)": 34.67,
+      "step": 6475,
+      "train_speed(iter/s)": 2.391213
+    },
+    {
+      "acc": 0.89854393,
+      "epoch": 11.696122633002705,
+      "grad_norm": 3.8378194348058896,
+      "learning_rate": 4.017646740097904e-06,
+      "loss": 0.62174253,
+      "memory(GiB)": 34.67,
+      "step": 6480,
+      "train_speed(iter/s)": 2.367355
+    },
+    {
+      "acc": 0.87098484,
+      "epoch": 11.705139765554554,
+      "grad_norm": 4.875289121619593,
+      "learning_rate": 4.010332668309936e-06,
+      "loss": 0.71115298,
+      "memory(GiB)": 34.67,
+      "step": 6485,
+      "train_speed(iter/s)": 2.343979
+    },
+    {
+      "acc": 0.88149166,
+      "epoch": 11.714156898106403,
+      "grad_norm": 3.612520836313393,
+      "learning_rate": 4.003020801587679e-06,
+      "loss": 0.60402894,
+      "memory(GiB)": 34.67,
+      "step": 6490,
+      "train_speed(iter/s)": 2.321111
+    },
+    {
+      "acc": 0.88029213,
+      "epoch": 11.723174030658251,
+      "grad_norm": 2.8555388407255604,
+      "learning_rate": 3.995711156214386e-06,
+      "loss": 0.63368893,
+      "memory(GiB)": 34.67,
+      "step": 6495,
+      "train_speed(iter/s)": 2.29874
+    },
+    {
+      "acc": 0.89712105,
+      "epoch": 11.7321911632101,
+      "grad_norm": 3.4431630187392335,
+      "learning_rate": 3.9884037484683654e-06,
+      "loss": 0.49287953,
+      "memory(GiB)": 34.67,
+      "step": 6500,
+      "train_speed(iter/s)": 2.276833
+    },
+    {
+      "acc": 0.89483719,
+      "epoch": 11.741208295761947,
+      "grad_norm": 3.116246226804721,
+      "learning_rate": 3.981098594622941e-06,
+      "loss": 0.57070303,
+      "memory(GiB)": 34.67,
+      "step": 6505,
+      "train_speed(iter/s)": 2.255368
+    },
+    {
+      "acc": 0.89828339,
+      "epoch": 11.750225428313795,
+      "grad_norm": 4.168551371962979,
+      "learning_rate": 3.973795710946418e-06,
+      "loss": 0.58798347,
+      "memory(GiB)": 34.67,
+      "step": 6510,
+      "train_speed(iter/s)": 2.234309
+    },
+    {
+      "acc": 0.8866642,
+      "epoch": 11.759242560865644,
+      "grad_norm": 5.510974999962866,
+      "learning_rate": 3.966495113702046e-06,
+      "loss": 0.60468707,
+      "memory(GiB)": 34.67,
+      "step": 6515,
+      "train_speed(iter/s)": 2.213712
+    },
+    {
+      "acc": 0.90433388,
+      "epoch": 11.768259693417493,
+      "grad_norm": 3.700239501504165,
+      "learning_rate": 3.959196819147983e-06,
+      "loss": 0.56499805,
+      "memory(GiB)": 34.67,
+      "step": 6520,
+      "train_speed(iter/s)": 2.193534
+    },
+    {
+      "acc": 0.89157143,
+      "epoch": 11.777276825969341,
+      "grad_norm": 2.9027667808878728,
+      "learning_rate": 3.951900843537258e-06,
+      "loss": 0.55809302,
+      "memory(GiB)": 34.67,
+      "step": 6525,
+      "train_speed(iter/s)": 2.173749
+    },
+    {
+      "acc": 0.88304062,
+      "epoch": 11.78629395852119,
+      "grad_norm": 2.7546405080588374,
+      "learning_rate": 3.944607203117736e-06,
+      "loss": 0.60958209,
+      "memory(GiB)": 34.67,
+      "step": 6530,
+      "train_speed(iter/s)": 2.154292
+    },
+    {
+      "acc": 0.89274206,
+      "epoch": 11.795311091073039,
+      "grad_norm": 2.9457983338903353,
+      "learning_rate": 3.937315914132086e-06,
+      "loss": 0.6001379,
+      "memory(GiB)": 34.67,
+      "step": 6535,
+      "train_speed(iter/s)": 2.13522
+    },
+    {
+      "acc": 0.88494377,
+      "epoch": 11.804328223624887,
+      "grad_norm": 3.08912343148247,
+      "learning_rate": 3.930026992817731e-06,
+      "loss": 0.62496905,
+      "memory(GiB)": 34.67,
+      "step": 6540,
+      "train_speed(iter/s)": 2.116564
+    },
+    {
+      "acc": 0.901404,
+      "epoch": 11.813345356176736,
+      "grad_norm": 2.7543128788149156,
+      "learning_rate": 3.922740455406831e-06,
+      "loss": 0.53852654,
+      "memory(GiB)": 34.67,
+      "step": 6545,
+      "train_speed(iter/s)": 2.09826
+    },
+    {
+      "acc": 0.87935314,
+      "epoch": 11.822362488728585,
+      "grad_norm": 3.011957805835299,
+      "learning_rate": 3.915456318126231e-06,
+      "loss": 0.64246588,
+      "memory(GiB)": 34.67,
+      "step": 6550,
+      "train_speed(iter/s)": 2.080237
+    },
+    {
+      "acc": 0.90037899,
+      "epoch": 11.831379621280433,
+      "grad_norm": 2.303245000726657,
+      "learning_rate": 3.908174597197433e-06,
+      "loss": 0.56113682,
+      "memory(GiB)": 34.67,
+      "step": 6555,
+      "train_speed(iter/s)": 2.062574
+    },
+    {
+      "acc": 0.89349146,
+      "epoch": 11.840396753832282,
+      "grad_norm": 2.946873472049538,
+      "learning_rate": 3.900895308836558e-06,
+      "loss": 0.58248949,
+      "memory(GiB)": 34.67,
+      "step": 6560,
+      "train_speed(iter/s)": 2.045216
+    },
+    {
+      "acc": 0.91550884,
+      "epoch": 11.84941388638413,
+      "grad_norm": 4.106085417976899,
+      "learning_rate": 3.893618469254311e-06,
+      "loss": 0.48115005,
+      "memory(GiB)": 34.67,
+      "step": 6565,
+      "train_speed(iter/s)": 2.02817
+    },
+    {
+      "acc": 0.89332514,
+      "epoch": 11.858431018935978,
+      "grad_norm": 4.238738103103749,
+      "learning_rate": 3.886344094655941e-06,
+      "loss": 0.56668701,
+      "memory(GiB)": 34.67,
+      "step": 6570,
+      "train_speed(iter/s)": 2.011458
+    },
+    {
+      "acc": 0.90490742,
+      "epoch": 11.867448151487826,
+      "grad_norm": 3.126016628257221,
+      "learning_rate": 3.879072201241209e-06,
+      "loss": 0.51700487,
+      "memory(GiB)": 34.67,
+      "step": 6575,
+      "train_speed(iter/s)": 1.995037
+    },
+    {
+      "acc": 0.89984188,
+      "epoch": 11.876465284039675,
+      "grad_norm": 1.841121350103198,
+      "learning_rate": 3.871802805204352e-06,
+      "loss": 0.53324027,
+      "memory(GiB)": 34.67,
+      "step": 6580,
+      "train_speed(iter/s)": 1.978906
+    },
+    {
+      "acc": 0.8948185,
+      "epoch": 11.885482416591524,
+      "grad_norm": 4.011181853907612,
+      "learning_rate": 3.86453592273404e-06,
+      "loss": 0.58295531,
+      "memory(GiB)": 34.67,
+      "step": 6585,
+      "train_speed(iter/s)": 1.963059
+    },
+    {
+      "acc": 0.90139599,
+      "epoch": 11.894499549143372,
+      "grad_norm": 5.673338290492078,
+      "learning_rate": 3.857271570013353e-06,
+      "loss": 0.55692811,
+      "memory(GiB)": 34.67,
+      "step": 6590,
+      "train_speed(iter/s)": 1.947502
+    },
+    {
+      "acc": 0.8901413,
+      "epoch": 11.903516681695221,
+      "grad_norm": 4.086904198177409,
+      "learning_rate": 3.8500097632197326e-06,
+      "loss": 0.57388477,
+      "memory(GiB)": 34.67,
+      "step": 6595,
+      "train_speed(iter/s)": 1.932209
+    },
+    {
+      "acc": 0.90845718,
+      "epoch": 11.91253381424707,
+      "grad_norm": 3.093457960696665,
+      "learning_rate": 3.842750518524949e-06,
+      "loss": 0.49866705,
+      "memory(GiB)": 34.67,
+      "step": 6600,
+      "train_speed(iter/s)": 1.917176
+    },
+    {
+      "acc": 0.89074831,
+      "epoch": 11.921550946798918,
+      "grad_norm": 4.125866495526902,
+      "learning_rate": 3.835493852095073e-06,
+      "loss": 0.55677662,
+      "memory(GiB)": 34.67,
+      "step": 6605,
+      "train_speed(iter/s)": 1.902415
+    },
+    {
+      "acc": 0.88955154,
+      "epoch": 11.930568079350767,
+      "grad_norm": 2.0918888964564273,
+      "learning_rate": 3.8282397800904275e-06,
+      "loss": 0.57829523,
+      "memory(GiB)": 34.67,
+      "step": 6610,
+      "train_speed(iter/s)": 1.887899
+    },
+    {
+      "acc": 0.88710747,
+      "epoch": 11.939585211902616,
+      "grad_norm": 3.713753590744293,
+      "learning_rate": 3.820988318665563e-06,
+      "loss": 0.597786,
+      "memory(GiB)": 34.67,
+      "step": 6615,
+      "train_speed(iter/s)": 1.873629
+    },
+    {
+      "acc": 0.89293966,
+      "epoch": 11.948602344454464,
+      "grad_norm": 3.392903276630068,
+      "learning_rate": 3.81373948396921e-06,
+      "loss": 0.54176407,
+      "memory(GiB)": 34.67,
+      "step": 6620,
+      "train_speed(iter/s)": 1.859592
+    },
+    {
+      "acc": 0.89268913,
+      "epoch": 11.957619477006311,
+      "grad_norm": 2.4845625647116334,
+      "learning_rate": 3.8064932921442553e-06,
+      "loss": 0.57533016,
+      "memory(GiB)": 34.67,
+      "step": 6625,
+      "train_speed(iter/s)": 1.845791
+    },
+    {
+      "acc": 0.89840698,
+      "epoch": 11.96663660955816,
+      "grad_norm": 2.7133497223785823,
+      "learning_rate": 3.799249759327697e-06,
+      "loss": 0.54027977,
+      "memory(GiB)": 34.67,
+      "step": 6630,
+      "train_speed(iter/s)": 1.831589
+    },
+    {
+      "acc": 0.88528519,
+      "epoch": 11.975653742110008,
+      "grad_norm": 4.907637284645984,
+      "learning_rate": 3.792008901650613e-06,
+      "loss": 0.6101069,
+      "memory(GiB)": 34.67,
+      "step": 6635,
+      "train_speed(iter/s)": 1.818238
+    },
+    {
+      "acc": 0.90314674,
+      "epoch": 11.984670874661857,
+      "grad_norm": 3.494447524122149,
+      "learning_rate": 3.7847707352381213e-06,
+      "loss": 0.50569868,
+      "memory(GiB)": 34.67,
+      "step": 6640,
+      "train_speed(iter/s)": 1.805082
+    },
+    {
+      "acc": 0.87980185,
+      "epoch": 11.993688007213706,
+      "grad_norm": 5.334889538473327,
+      "learning_rate": 3.777535276209352e-06,
+      "loss": 0.64969778,
+      "memory(GiB)": 34.67,
+      "step": 6645,
+      "train_speed(iter/s)": 1.792159
+    },
+    {
+      "epoch": 12.0,
+      "eval_acc": 0.8994181213632585,
+      "eval_loss": 0.2734980285167694,
+      "eval_runtime": 50.0068,
+      "eval_samples_per_second": 9.339,
+      "eval_steps_per_second": 1.18,
+      "step": 6649
+    },
+    {
+      "acc": 0.81350288,
+      "epoch": 12.001803426510369,
+      "grad_norm": 2.575214669644955,
+      "learning_rate": 3.770302540677399e-06,
+      "loss": 0.45334911,
+      "memory(GiB)": 34.67,
+      "step": 6650,
+      "train_speed(iter/s)": 1.733284
+    },
+    {
+      "acc": 0.90637493,
+      "epoch": 12.010820559062218,
+      "grad_norm": 3.4925685024652893,
+      "learning_rate": 3.763072544749297e-06,
+      "loss": 0.51629629,
+      "memory(GiB)": 34.67,
+      "step": 6655,
+      "train_speed(iter/s)": 1.721439
+    },
+    {
+      "acc": 0.92434101,
+      "epoch": 12.019837691614066,
+      "grad_norm": 2.039044261617016,
+      "learning_rate": 3.7558453045259762e-06,
+      "loss": 0.42532778,
+      "memory(GiB)": 34.67,
+      "step": 6660,
+      "train_speed(iter/s)": 1.709787
+    },
+    {
+      "acc": 0.89824448,
+      "epoch": 12.028854824165915,
+      "grad_norm": 2.1155265660242306,
+      "learning_rate": 3.74862083610223e-06,
+      "loss": 0.54138293,
+      "memory(GiB)": 34.67,
+      "step": 6665,
+      "train_speed(iter/s)": 1.698271
+    },
+    {
+      "acc": 0.91347828,
+      "epoch": 12.037871956717764,
+      "grad_norm": 3.418855024878286,
+      "learning_rate": 3.74139915556668e-06,
+      "loss": 0.48412566,
+      "memory(GiB)": 34.67,
+      "step": 6670,
+      "train_speed(iter/s)": 1.68692
+    },
+    {
+      "acc": 0.89754686,
+      "epoch": 12.046889089269612,
+      "grad_norm": 3.3779581181312692,
+      "learning_rate": 3.7341802790017427e-06,
+      "loss": 0.5130682,
+      "memory(GiB)": 34.67,
+      "step": 6675,
+      "train_speed(iter/s)": 1.675744
+    },
+    {
+      "acc": 0.91587677,
+      "epoch": 12.055906221821461,
+      "grad_norm": 5.789073567685528,
+      "learning_rate": 3.7269642224835843e-06,
+      "loss": 0.44497404,
+      "memory(GiB)": 34.67,
+      "step": 6680,
+      "train_speed(iter/s)": 1.664737
+    },
+    {
+      "acc": 0.89080715,
+      "epoch": 12.06492335437331,
+      "grad_norm": 3.831420518156388,
+      "learning_rate": 3.719751002082094e-06,
+      "loss": 0.55942383,
+      "memory(GiB)": 34.67,
+      "step": 6685,
+      "train_speed(iter/s)": 1.65391
+    },
+    {
+      "acc": 0.91208267,
+      "epoch": 12.073940486925158,
+      "grad_norm": 5.172936670113741,
+      "learning_rate": 3.7125406338608438e-06,
+      "loss": 0.44970398,
+      "memory(GiB)": 34.67,
+      "step": 6690,
+      "train_speed(iter/s)": 1.643196
+    },
+    {
+      "acc": 0.92443676,
+      "epoch": 12.082957619477007,
+      "grad_norm": 2.8504321431159267,
+      "learning_rate": 3.7053331338770557e-06,
+      "loss": 0.40664091,
+      "memory(GiB)": 34.67,
+      "step": 6695,
+      "train_speed(iter/s)": 1.632677
+    },
+    {
+      "acc": 0.91076431,
+      "epoch": 12.091974752028856,
+      "grad_norm": 3.604219094120456,
+      "learning_rate": 3.698128518181563e-06,
+      "loss": 0.45517664,
+      "memory(GiB)": 34.67,
+      "step": 6700,
+      "train_speed(iter/s)": 1.622263
+    },
+    {
+      "acc": 0.89761791,
+      "epoch": 12.100991884580703,
+      "grad_norm": 3.6072006583559166,
+      "learning_rate": 3.6909268028187756e-06,
+      "loss": 0.51308622,
+      "memory(GiB)": 34.67,
+      "step": 6705,
+      "train_speed(iter/s)": 1.612019
+    },
+    {
+      "acc": 0.90031252,
+      "epoch": 12.110009017132551,
+      "grad_norm": 3.28548310243231,
+      "learning_rate": 3.6837280038266453e-06,
+      "loss": 0.52073536,
+      "memory(GiB)": 34.67,
+      "step": 6710,
+      "train_speed(iter/s)": 1.601921
+    },
+    {
+      "acc": 0.88845291,
+      "epoch": 12.1190261496844,
+      "grad_norm": 2.7131167802900142,
+      "learning_rate": 3.6765321372366295e-06,
+      "loss": 0.58866653,
+      "memory(GiB)": 34.67,
+      "step": 6715,
+      "train_speed(iter/s)": 1.591964
+    },
+    {
+      "acc": 0.89709702,
+      "epoch": 12.128043282236249,
+      "grad_norm": 3.324104635839056,
+      "learning_rate": 3.669339219073653e-06,
+      "loss": 0.5729095,
+      "memory(GiB)": 34.67,
+      "step": 6720,
+      "train_speed(iter/s)": 1.582143
+    },
+    {
+      "acc": 0.90601044,
+      "epoch": 12.137060414788097,
+      "grad_norm": 3.9513577185198887,
+      "learning_rate": 3.6621492653560786e-06,
+      "loss": 0.51798992,
+      "memory(GiB)": 34.67,
+      "step": 6725,
+      "train_speed(iter/s)": 1.572461
+    },
+    {
+      "acc": 0.90220985,
+      "epoch": 12.146077547339946,
+      "grad_norm": 2.743185886658063,
+      "learning_rate": 3.654962292095663e-06,
+      "loss": 0.53028731,
+      "memory(GiB)": 34.67,
+      "step": 6730,
+      "train_speed(iter/s)": 1.562919
+    },
+    {
+      "acc": 0.90385151,
+      "epoch": 12.155094679891794,
+      "grad_norm": 3.8011841848228323,
+      "learning_rate": 3.647778315297528e-06,
+      "loss": 0.49340973,
+      "memory(GiB)": 34.67,
+      "step": 6735,
+      "train_speed(iter/s)": 1.553487
+    },
+    {
+      "acc": 0.89563923,
+      "epoch": 12.164111812443643,
+      "grad_norm": 4.042298265702815,
+      "learning_rate": 3.6405973509601235e-06,
+      "loss": 0.57576737,
+      "memory(GiB)": 34.67,
+      "step": 6740,
+      "train_speed(iter/s)": 1.544179
+    },
+    {
+      "acc": 0.90624695,
+      "epoch": 12.173128944995492,
+      "grad_norm": 2.2936041552366215,
+      "learning_rate": 3.633419415075186e-06,
+      "loss": 0.52801132,
+      "memory(GiB)": 34.67,
+      "step": 6745,
+      "train_speed(iter/s)": 1.534999
+    },
+    {
+      "acc": 0.89541893,
+      "epoch": 12.18214607754734,
+      "grad_norm": 2.0906405577820473,
+      "learning_rate": 3.6262445236277153e-06,
+      "loss": 0.55127258,
+      "memory(GiB)": 34.67,
+      "step": 6750,
+      "train_speed(iter/s)": 1.525961
+    },
+    {
+      "acc": 0.91594315,
+      "epoch": 12.19116321009919,
+      "grad_norm": 2.1409635042593913,
+      "learning_rate": 3.6190726925959222e-06,
+      "loss": 0.43182116,
+      "memory(GiB)": 34.67,
+      "step": 6755,
+      "train_speed(iter/s)": 1.517033
+    },
+    {
+      "acc": 0.90597992,
+      "epoch": 12.200180342651038,
+      "grad_norm": 2.926711327961975,
+      "learning_rate": 3.6119039379512115e-06,
+      "loss": 0.47865562,
+      "memory(GiB)": 34.67,
+      "step": 6760,
+      "train_speed(iter/s)": 1.508237
+    },
+    {
+      "acc": 0.90510683,
+      "epoch": 12.209197475202885,
+      "grad_norm": 2.604994122390104,
+      "learning_rate": 3.6047382756581285e-06,
+      "loss": 0.53902903,
+      "memory(GiB)": 34.67,
+      "step": 6765,
+      "train_speed(iter/s)": 1.499534
+    },
+    {
+      "acc": 0.91268291,
+      "epoch": 12.218214607754733,
+      "grad_norm": 1.9489388242574992,
+      "learning_rate": 3.5975757216743377e-06,
+      "loss": 0.42435141,
+      "memory(GiB)": 34.67,
+      "step": 6770,
+      "train_speed(iter/s)": 1.490931
+    },
+    {
+      "acc": 0.89771376,
+      "epoch": 12.227231740306582,
+      "grad_norm": 2.997573524241079,
+      "learning_rate": 3.5904162919505788e-06,
+      "loss": 0.57115669,
+      "memory(GiB)": 34.67,
+      "step": 6775,
+      "train_speed(iter/s)": 1.482461
+    },
+    {
+      "acc": 0.92218046,
+      "epoch": 12.23624887285843,
+      "grad_norm": 2.4411178735663697,
+      "learning_rate": 3.5832600024306345e-06,
+      "loss": 0.41223426,
+      "memory(GiB)": 34.67,
+      "step": 6780,
+      "train_speed(iter/s)": 1.474083
+    },
+    {
+      "acc": 0.88631496,
+      "epoch": 12.24526600541028,
+      "grad_norm": 4.293779878220328,
+      "learning_rate": 3.5761068690512978e-06,
+      "loss": 0.61934638,
+      "memory(GiB)": 34.67,
+      "step": 6785,
+      "train_speed(iter/s)": 1.465822
+    },
+    {
+      "acc": 0.90367622,
+      "epoch": 12.254283137962128,
+      "grad_norm": 2.0034139313461417,
+      "learning_rate": 3.5689569077423262e-06,
+      "loss": 0.51608496,
+      "memory(GiB)": 34.67,
+      "step": 6790,
+      "train_speed(iter/s)": 1.457684
+    },
+    {
+      "acc": 0.9154211,
+      "epoch": 12.263300270513977,
+      "grad_norm": 3.3568476733945207,
+      "learning_rate": 3.5618101344264176e-06,
+      "loss": 0.4631865,
+      "memory(GiB)": 34.67,
+      "step": 6795,
+      "train_speed(iter/s)": 1.449617
+    },
+    {
+      "acc": 0.90154781,
+      "epoch": 12.272317403065825,
+      "grad_norm": 2.1888321459224094,
+      "learning_rate": 3.5546665650191707e-06,
+      "loss": 0.51787224,
+      "memory(GiB)": 34.67,
+      "step": 6800,
+      "train_speed(iter/s)": 1.441671
+    },
+    {
+      "acc": 0.90700722,
+      "epoch": 12.281334535617674,
+      "grad_norm": 2.292007057868882,
+      "learning_rate": 3.5475262154290474e-06,
+      "loss": 0.45840693,
+      "memory(GiB)": 34.67,
+      "step": 6805,
+      "train_speed(iter/s)": 1.433802
+    },
+    {
+      "acc": 0.89161024,
+      "epoch": 12.290351668169523,
+      "grad_norm": 2.8701291268559577,
+      "learning_rate": 3.5403891015573407e-06,
+      "loss": 0.56402502,
+      "memory(GiB)": 34.67,
+      "step": 6810,
+      "train_speed(iter/s)": 1.426029
+    },
+    {
+      "acc": 0.90689182,
+      "epoch": 12.299368800721371,
+      "grad_norm": 3.4158428604143443,
+      "learning_rate": 3.533255239298135e-06,
+      "loss": 0.48761907,
+      "memory(GiB)": 34.67,
+      "step": 6815,
+      "train_speed(iter/s)": 1.418349
+    },
+    {
+      "acc": 0.88538704,
+      "epoch": 12.30838593327322,
+      "grad_norm": 2.744623166849405,
+      "learning_rate": 3.5261246445382768e-06,
+      "loss": 0.59895587,
+      "memory(GiB)": 34.67,
+      "step": 6820,
+      "train_speed(iter/s)": 1.410775
+    },
+    {
+      "acc": 0.90757618,
+      "epoch": 12.317403065825067,
+      "grad_norm": 2.348826790027136,
+      "learning_rate": 3.5189973331573347e-06,
+      "loss": 0.49169731,
+      "memory(GiB)": 34.67,
+      "step": 6825,
+      "train_speed(iter/s)": 1.403295
+    },
+    {
+      "acc": 0.90024757,
+      "epoch": 12.326420198376915,
+      "grad_norm": 3.4637134352625245,
+      "learning_rate": 3.5118733210275653e-06,
+      "loss": 0.53719163,
+      "memory(GiB)": 34.67,
+      "step": 6830,
+      "train_speed(iter/s)": 1.395908
+    },
+    {
+      "acc": 0.90511379,
+      "epoch": 12.335437330928764,
+      "grad_norm": 3.1429330020547877,
+      "learning_rate": 3.5047526240138756e-06,
+      "loss": 0.48838396,
+      "memory(GiB)": 34.67,
+      "step": 6835,
+      "train_speed(iter/s)": 1.388601
+    },
+    {
+      "acc": 0.89170218,
+      "epoch": 12.344454463480613,
+      "grad_norm": 2.77515228069174,
+      "learning_rate": 3.4976352579737966e-06,
+      "loss": 0.56038818,
+      "memory(GiB)": 34.67,
+      "step": 6840,
+      "train_speed(iter/s)": 1.381386
+    },
+    {
+      "acc": 0.88870592,
+      "epoch": 12.353471596032461,
+      "grad_norm": 2.612572066007513,
+      "learning_rate": 3.490521238757432e-06,
+      "loss": 0.58700676,
+      "memory(GiB)": 34.67,
+      "step": 6845,
+      "train_speed(iter/s)": 1.374257
+    },
+    {
+      "acc": 0.89043579,
+      "epoch": 12.36248872858431,
+      "grad_norm": 2.0526368361926117,
+      "learning_rate": 3.483410582207442e-06,
+      "loss": 0.56371317,
+      "memory(GiB)": 34.67,
+      "step": 6850,
+      "train_speed(iter/s)": 1.366813
+    },
+    {
+      "acc": 0.9151659,
+      "epoch": 12.371505861136159,
+      "grad_norm": 2.4417612779693547,
+      "learning_rate": 3.4763033041589896e-06,
+      "loss": 0.43916836,
+      "memory(GiB)": 34.67,
+      "step": 6855,
+      "train_speed(iter/s)": 1.359454
+    },
+    {
+      "acc": 0.91412525,
+      "epoch": 12.380522993688007,
+      "grad_norm": 3.774629777758544,
+      "learning_rate": 3.4691994204397204e-06,
+      "loss": 0.45177279,
+      "memory(GiB)": 34.67,
+      "step": 6860,
+      "train_speed(iter/s)": 1.352595
+    },
+    {
+      "acc": 0.90705338,
+      "epoch": 12.389540126239856,
+      "grad_norm": 3.0208576391297237,
+      "learning_rate": 3.4620989468697165e-06,
+      "loss": 0.49362307,
+      "memory(GiB)": 34.67,
+      "step": 6865,
+      "train_speed(iter/s)": 1.345799
+    },
+    {
+      "acc": 0.92138443,
+      "epoch": 12.398557258791705,
+      "grad_norm": 2.6168496358492157,
+      "learning_rate": 3.455001899261469e-06,
+      "loss": 0.47534761,
+      "memory(GiB)": 34.67,
+      "step": 6870,
+      "train_speed(iter/s)": 1.339063
+    },
+    {
+      "acc": 0.9038641,
+      "epoch": 12.407574391343553,
+      "grad_norm": 2.8679557193192022,
+      "learning_rate": 3.4479082934198378e-06,
+      "loss": 0.50212803,
+      "memory(GiB)": 34.67,
+      "step": 6875,
+      "train_speed(iter/s)": 1.332421
+    },
+    {
+      "acc": 0.9250123,
+      "epoch": 12.416591523895402,
+      "grad_norm": 3.3074929546790988,
+      "learning_rate": 3.4408181451420184e-06,
+      "loss": 0.40797939,
+      "memory(GiB)": 34.67,
+      "step": 6880,
+      "train_speed(iter/s)": 1.325851
+    },
+    {
+      "acc": 0.87166634,
+      "epoch": 12.425608656447249,
+      "grad_norm": 2.8634096224537005,
+      "learning_rate": 3.4337314702175056e-06,
+      "loss": 0.62051229,
+      "memory(GiB)": 34.67,
+      "step": 6885,
+      "train_speed(iter/s)": 1.319368
+    },
+    {
+      "acc": 0.89600964,
+      "epoch": 12.434625788999098,
+      "grad_norm": 3.5263444180073056,
+      "learning_rate": 3.426648284428063e-06,
+      "loss": 0.55068827,
+      "memory(GiB)": 34.67,
+      "step": 6890,
+      "train_speed(iter/s)": 1.312945
+    },
+    {
+      "acc": 0.90047712,
+      "epoch": 12.443642921550946,
+      "grad_norm": 2.5930249467829225,
+      "learning_rate": 3.419568603547678e-06,
+      "loss": 0.49448376,
+      "memory(GiB)": 34.67,
+      "step": 6895,
+      "train_speed(iter/s)": 1.306577
+    },
+    {
+      "acc": 0.92914753,
+      "epoch": 12.452660054102795,
+      "grad_norm": 2.24287860956484,
+      "learning_rate": 3.412492443342538e-06,
+      "loss": 0.39698575,
+      "memory(GiB)": 34.67,
+      "step": 6900,
+      "train_speed(iter/s)": 1.300308
+    },
+    {
+      "acc": 0.88429451,
+      "epoch": 12.461677186654644,
+      "grad_norm": 4.0189329106208795,
+      "learning_rate": 3.405419819570986e-06,
+      "loss": 0.58602076,
+      "memory(GiB)": 34.67,
+      "step": 6905,
+      "train_speed(iter/s)": 1.294077
+    },
+    {
+      "acc": 0.89383793,
+      "epoch": 12.470694319206492,
+      "grad_norm": 4.231128706089549,
+      "learning_rate": 3.398350747983493e-06,
+      "loss": 0.57240362,
+      "memory(GiB)": 34.67,
+      "step": 6910,
+      "train_speed(iter/s)": 1.287912
+    },
+    {
+      "acc": 0.91105804,
+      "epoch": 12.479711451758341,
+      "grad_norm": 2.676913257049845,
+      "learning_rate": 3.391285244322616e-06,
+      "loss": 0.47283764,
+      "memory(GiB)": 34.67,
+      "step": 6915,
+      "train_speed(iter/s)": 1.281821
+    },
+    {
+      "acc": 0.9111577,
+      "epoch": 12.48872858431019,
+      "grad_norm": 1.7914679032768803,
+      "learning_rate": 3.38422332432297e-06,
+      "loss": 0.4607234,
+      "memory(GiB)": 34.67,
+      "step": 6920,
+      "train_speed(iter/s)": 1.275814
+    },
+    {
+      "acc": 0.8947032,
+      "epoch": 12.497745716862038,
+      "grad_norm": 2.08201057362077,
+      "learning_rate": 3.377165003711185e-06,
+      "loss": 0.55225325,
+      "memory(GiB)": 34.67,
+      "step": 6925,
+      "train_speed(iter/s)": 1.269869
+    },
+    {
+      "acc": 0.90196323,
+      "epoch": 12.506762849413887,
+      "grad_norm": 2.3669301135549947,
+      "learning_rate": 3.37011029820588e-06,
+      "loss": 0.48145366,
+      "memory(GiB)": 34.67,
+      "step": 6930,
+      "train_speed(iter/s)": 1.263993
+    },
+    {
+      "acc": 0.88760262,
+      "epoch": 12.515779981965736,
+      "grad_norm": 3.5300593443343566,
+      "learning_rate": 3.36305922351762e-06,
+      "loss": 0.61123838,
+      "memory(GiB)": 34.67,
+      "step": 6935,
+      "train_speed(iter/s)": 1.258166
+    },
+    {
+      "acc": 0.89373741,
+      "epoch": 12.524797114517582,
+      "grad_norm": 2.9678437185920052,
+      "learning_rate": 3.3560117953488847e-06,
+      "loss": 0.54196587,
+      "memory(GiB)": 34.67,
+      "step": 6940,
+      "train_speed(iter/s)": 1.252408
+    },
+    {
+      "acc": 0.90031862,
+      "epoch": 12.533814247069431,
+      "grad_norm": 2.9632955771816594,
+      "learning_rate": 3.348968029394035e-06,
+      "loss": 0.52354126,
+      "memory(GiB)": 34.67,
+      "step": 6945,
+      "train_speed(iter/s)": 1.246707
+    },
+    {
+      "acc": 0.91497507,
+      "epoch": 12.54283137962128,
+      "grad_norm": 3.8620935294366237,
+      "learning_rate": 3.3419279413392737e-06,
+      "loss": 0.45280676,
+      "memory(GiB)": 34.67,
+      "step": 6950,
+      "train_speed(iter/s)": 1.241069
+    },
+    {
+      "acc": 0.90571871,
+      "epoch": 12.551848512173128,
+      "grad_norm": 2.2996013995351725,
+      "learning_rate": 3.334891546862616e-06,
+      "loss": 0.5276722,
+      "memory(GiB)": 34.67,
+      "step": 6955,
+      "train_speed(iter/s)": 1.235494
+    },
+    {
+      "acc": 0.90228748,
+      "epoch": 12.560865644724977,
+      "grad_norm": 2.2964152470744357,
+      "learning_rate": 3.327858861633847e-06,
+      "loss": 0.5086122,
+      "memory(GiB)": 34.67,
+      "step": 6960,
+      "train_speed(iter/s)": 1.229959
+    },
+    {
+      "acc": 0.91079998,
+      "epoch": 12.569882777276826,
+      "grad_norm": 3.6548937928907845,
+      "learning_rate": 3.3208299013144974e-06,
+      "loss": 0.49442649,
+      "memory(GiB)": 34.67,
+      "step": 6965,
+      "train_speed(iter/s)": 1.224506
+    },
+    {
+      "acc": 0.89158049,
+      "epoch": 12.578899909828674,
+      "grad_norm": 2.8246531241605957,
+      "learning_rate": 3.3138046815577973e-06,
+      "loss": 0.55396352,
+      "memory(GiB)": 34.67,
+      "step": 6970,
+      "train_speed(iter/s)": 1.219097
+    },
+    {
+      "acc": 0.90056505,
+      "epoch": 12.587917042380523,
+      "grad_norm": 2.947643468854448,
+      "learning_rate": 3.3067832180086513e-06,
+      "loss": 0.52505016,
+      "memory(GiB)": 34.67,
+      "step": 6975,
+      "train_speed(iter/s)": 1.213741
+    },
+    {
+      "acc": 0.90709219,
+      "epoch": 12.596934174932372,
+      "grad_norm": 2.872902939668548,
+      "learning_rate": 3.2997655263035942e-06,
+      "loss": 0.49564762,
+      "memory(GiB)": 34.67,
+      "step": 6980,
+      "train_speed(iter/s)": 1.208446
+    },
+    {
+      "acc": 0.91496458,
+      "epoch": 12.60595130748422,
+      "grad_norm": 2.7179826260252256,
+      "learning_rate": 3.2927516220707667e-06,
+      "loss": 0.46064801,
+      "memory(GiB)": 34.67,
+      "step": 6985,
+      "train_speed(iter/s)": 1.203193
+    },
+    {
+      "acc": 0.89439774,
+      "epoch": 12.614968440036069,
+      "grad_norm": 2.5661195393731386,
+      "learning_rate": 3.285741520929869e-06,
+      "loss": 0.52741041,
+      "memory(GiB)": 34.67,
+      "step": 6990,
+      "train_speed(iter/s)": 1.197989
+    },
+    {
+      "acc": 0.91784515,
+      "epoch": 12.623985572587918,
+      "grad_norm": 3.036083624872931,
+      "learning_rate": 3.2787352384921335e-06,
+      "loss": 0.45988255,
+      "memory(GiB)": 34.67,
+      "step": 6995,
+      "train_speed(iter/s)": 1.192842
+    },
+    {
+      "acc": 0.89744816,
+      "epoch": 12.633002705139766,
+      "grad_norm": 2.6472984324447726,
+      "learning_rate": 3.271732790360295e-06,
+      "loss": 0.500772,
+      "memory(GiB)": 34.67,
+      "step": 7000,
+      "train_speed(iter/s)": 1.187755
+    },
+    {
+      "acc": 0.91127987,
+      "epoch": 12.642019837691613,
+      "grad_norm": 2.4186789506546047,
+      "learning_rate": 3.264734192128542e-06,
+      "loss": 0.45138502,
+      "memory(GiB)": 34.67,
+      "step": 7005,
+      "train_speed(iter/s)": 1.182709
+    },
+    {
+      "acc": 0.91185627,
+      "epoch": 12.651036970243462,
+      "grad_norm": 1.9607647456441446,
+      "learning_rate": 3.2577394593824894e-06,
+      "loss": 0.46712923,
+      "memory(GiB)": 34.67,
+      "step": 7010,
+      "train_speed(iter/s)": 1.177712
+    },
+    {
+      "acc": 0.90184002,
+      "epoch": 12.66005410279531,
+      "grad_norm": 2.9091787295472407,
+      "learning_rate": 3.2507486076991506e-06,
+      "loss": 0.53244228,
+      "memory(GiB)": 34.67,
+      "step": 7015,
+      "train_speed(iter/s)": 1.172769
+    },
+    {
+      "acc": 0.89695921,
+      "epoch": 12.66907123534716,
+      "grad_norm": 2.4308655171528293,
+      "learning_rate": 3.2437616526468867e-06,
+      "loss": 0.49879436,
+      "memory(GiB)": 34.67,
+      "step": 7020,
+      "train_speed(iter/s)": 1.167865
+    },
+    {
+      "acc": 0.91786423,
+      "epoch": 12.678088367899008,
+      "grad_norm": 2.071317989419375,
+      "learning_rate": 3.236778609785389e-06,
+      "loss": 0.45005484,
+      "memory(GiB)": 34.67,
+      "step": 7025,
+      "train_speed(iter/s)": 1.163017
+    },
+    {
+      "acc": 0.92142124,
+      "epoch": 12.687105500450857,
+      "grad_norm": 5.121700176171818,
+      "learning_rate": 3.2297994946656324e-06,
+      "loss": 0.44252462,
+      "memory(GiB)": 34.67,
+      "step": 7030,
+      "train_speed(iter/s)": 1.158207
+    },
+    {
+      "acc": 0.90308228,
+      "epoch": 12.696122633002705,
+      "grad_norm": 3.0897040621180554,
+      "learning_rate": 3.222824322829846e-06,
+      "loss": 0.52320132,
+      "memory(GiB)": 34.67,
+      "step": 7035,
+      "train_speed(iter/s)": 1.153452
+    },
+    {
+      "acc": 0.90803566,
+      "epoch": 12.705139765554554,
+      "grad_norm": 2.6214822502930555,
+      "learning_rate": 3.2158531098114775e-06,
+      "loss": 0.45875521,
+      "memory(GiB)": 34.67,
+      "step": 7040,
+      "train_speed(iter/s)": 1.148741
+    },
+    {
+      "acc": 0.93123388,
+      "epoch": 12.714156898106403,
+      "grad_norm": 2.761663377941063,
+      "learning_rate": 3.208885871135159e-06,
+      "loss": 0.36535487,
+      "memory(GiB)": 34.67,
+      "step": 7045,
+      "train_speed(iter/s)": 1.144068
+    },
+    {
+      "acc": 0.91526604,
+      "epoch": 12.723174030658251,
+      "grad_norm": 3.5710697229953556,
+      "learning_rate": 3.2019226223166682e-06,
+      "loss": 0.45317073,
+      "memory(GiB)": 34.67,
+      "step": 7050,
+      "train_speed(iter/s)": 1.139451
+    },
+    {
+      "acc": 0.90850372,
+      "epoch": 12.7321911632101,
+      "grad_norm": 3.0884367350937154,
+      "learning_rate": 3.194963378862904e-06,
+      "loss": 0.5030972,
+      "memory(GiB)": 34.67,
+      "step": 7055,
+      "train_speed(iter/s)": 1.134869
+    },
+    {
+      "acc": 0.89610424,
+      "epoch": 12.741208295761947,
+      "grad_norm": 2.479456654265497,
+      "learning_rate": 3.1880081562718397e-06,
+      "loss": 0.52796822,
+      "memory(GiB)": 34.67,
+      "step": 7060,
+      "train_speed(iter/s)": 1.130343
+    },
+    {
+      "acc": 0.92005167,
+      "epoch": 12.750225428313795,
+      "grad_norm": 1.9885855087772923,
+      "learning_rate": 3.1810569700324984e-06,
+      "loss": 0.4264123,
+      "memory(GiB)": 34.67,
+      "step": 7065,
+      "train_speed(iter/s)": 1.125857
+    },
+    {
+      "acc": 0.9267272,
+      "epoch": 12.759242560865644,
+      "grad_norm": 2.6471495841215082,
+      "learning_rate": 3.1741098356249096e-06,
+      "loss": 0.40731611,
+      "memory(GiB)": 34.67,
+      "step": 7070,
+      "train_speed(iter/s)": 1.12141
+    },
+    {
+      "acc": 0.91518879,
+      "epoch": 12.768259693417493,
+      "grad_norm": 1.9664739870788763,
+      "learning_rate": 3.167166768520085e-06,
+      "loss": 0.43133411,
+      "memory(GiB)": 34.67,
+      "step": 7075,
+      "train_speed(iter/s)": 1.116999
+    },
+    {
+      "acc": 0.90235577,
+      "epoch": 12.777276825969341,
+      "grad_norm": 3.236070038759275,
+      "learning_rate": 3.1602277841799765e-06,
+      "loss": 0.51252337,
+      "memory(GiB)": 34.67,
+      "step": 7080,
+      "train_speed(iter/s)": 1.112465
+    },
+    {
+      "acc": 0.90878696,
+      "epoch": 12.78629395852119,
+      "grad_norm": 2.531316943039322,
+      "learning_rate": 3.1532928980574405e-06,
+      "loss": 0.48896437,
+      "memory(GiB)": 34.67,
+      "step": 7085,
+      "train_speed(iter/s)": 1.108021
+    },
+    {
+      "acc": 0.90052872,
+      "epoch": 12.795311091073039,
+      "grad_norm": 2.6102614878041304,
+      "learning_rate": 3.1463621255962123e-06,
+      "loss": 0.56132298,
+      "memory(GiB)": 34.67,
+      "step": 7090,
+      "train_speed(iter/s)": 1.103742
+    },
+    {
+      "acc": 0.9197155,
+      "epoch": 12.804328223624887,
+      "grad_norm": 2.0247956504089393,
+      "learning_rate": 3.1394354822308628e-06,
+      "loss": 0.41660166,
+      "memory(GiB)": 34.67,
+      "step": 7095,
+      "train_speed(iter/s)": 1.099499
+    },
+    {
+      "acc": 0.91555653,
+      "epoch": 12.813345356176736,
+      "grad_norm": 2.5010295438945183,
+      "learning_rate": 3.1325129833867686e-06,
+      "loss": 0.45076404,
+      "memory(GiB)": 34.67,
+      "step": 7100,
+      "train_speed(iter/s)": 1.095289
+    },
+    {
+      "acc": 0.90210075,
+      "epoch": 12.822362488728585,
+      "grad_norm": 2.599043666518721,
+      "learning_rate": 3.125594644480074e-06,
+      "loss": 0.52382908,
+      "memory(GiB)": 34.67,
+      "step": 7105,
+      "train_speed(iter/s)": 1.091122
+    },
+    {
+      "acc": 0.90552273,
+      "epoch": 12.831379621280433,
+      "grad_norm": 2.1825629567923297,
+      "learning_rate": 3.118680480917664e-06,
+      "loss": 0.4995965,
+      "memory(GiB)": 34.67,
+      "step": 7110,
+      "train_speed(iter/s)": 1.086986
+    },
+    {
+      "acc": 0.90266972,
+      "epoch": 12.840396753832282,
+      "grad_norm": 3.4355316990646596,
+      "learning_rate": 3.111770508097125e-06,
+      "loss": 0.49682536,
+      "memory(GiB)": 34.67,
+      "step": 7115,
+      "train_speed(iter/s)": 1.082889
+    },
+    {
+      "acc": 0.90070534,
+      "epoch": 12.84941388638413,
+      "grad_norm": 2.604672639680177,
+      "learning_rate": 3.104864741406704e-06,
+      "loss": 0.48613911,
+      "memory(GiB)": 34.67,
+      "step": 7120,
+      "train_speed(iter/s)": 1.078822
+    },
+    {
+      "acc": 0.92008801,
+      "epoch": 12.858431018935978,
+      "grad_norm": 3.6092092728467233,
+      "learning_rate": 3.0979631962252906e-06,
+      "loss": 0.45216789,
+      "memory(GiB)": 34.67,
+      "step": 7125,
+      "train_speed(iter/s)": 1.074797
+    },
+    {
+      "acc": 0.90443535,
+      "epoch": 12.867448151487826,
+      "grad_norm": 2.791672773984579,
+      "learning_rate": 3.091065887922364e-06,
+      "loss": 0.50383472,
+      "memory(GiB)": 34.67,
+      "step": 7130,
+      "train_speed(iter/s)": 1.070818
+    },
+    {
+      "acc": 0.89264803,
+      "epoch": 12.876465284039675,
+      "grad_norm": 4.007330384306319,
+      "learning_rate": 3.0841728318579744e-06,
+      "loss": 0.55332375,
+      "memory(GiB)": 34.67,
+      "step": 7135,
+      "train_speed(iter/s)": 1.066878
+    },
+    {
+      "acc": 0.91145496,
+      "epoch": 12.885482416591524,
+      "grad_norm": 2.5195743939618462,
+      "learning_rate": 3.0772840433827026e-06,
+      "loss": 0.45711875,
+      "memory(GiB)": 34.67,
+      "step": 7140,
+      "train_speed(iter/s)": 1.062969
+    },
+    {
+      "acc": 0.90285149,
+      "epoch": 12.894499549143372,
+      "grad_norm": 4.04052031719566,
+      "learning_rate": 3.07039953783762e-06,
+      "loss": 0.47453213,
+      "memory(GiB)": 34.67,
+      "step": 7145,
+      "train_speed(iter/s)": 1.059088
+    },
+    {
+      "acc": 0.91616383,
+      "epoch": 12.903516681695221,
+      "grad_norm": 2.5113154832654367,
+      "learning_rate": 3.063519330554266e-06,
+      "loss": 0.45489578,
+      "memory(GiB)": 34.67,
+      "step": 7150,
+      "train_speed(iter/s)": 1.055244
+    },
+    {
+      "acc": 0.89737759,
+      "epoch": 12.91253381424707,
+      "grad_norm": 2.211501042980457,
+      "learning_rate": 3.0566434368546043e-06,
+      "loss": 0.5210609,
+      "memory(GiB)": 34.67,
+      "step": 7155,
+      "train_speed(iter/s)": 1.051417
+    },
+    {
+      "acc": 0.90460377,
+      "epoch": 12.921550946798918,
+      "grad_norm": 3.1140805661653372,
+      "learning_rate": 3.049771872050995e-06,
+      "loss": 0.50720844,
+      "memory(GiB)": 34.67,
+      "step": 7160,
+      "train_speed(iter/s)": 1.047636
+    },
+    {
+      "acc": 0.90584793,
+      "epoch": 12.930568079350767,
+      "grad_norm": 3.407546548366909,
+      "learning_rate": 3.042904651446154e-06,
+      "loss": 0.48453746,
+      "memory(GiB)": 34.67,
+      "step": 7165,
+      "train_speed(iter/s)": 1.043889
+    },
+    {
+      "acc": 0.91362514,
+      "epoch": 12.939585211902616,
+      "grad_norm": 2.5606100097716453,
+      "learning_rate": 3.036041790333129e-06,
+      "loss": 0.48806229,
+      "memory(GiB)": 34.67,
+      "step": 7170,
+      "train_speed(iter/s)": 1.04017
+    },
+    {
+      "acc": 0.90656824,
+      "epoch": 12.948602344454464,
+      "grad_norm": 2.405704397584747,
+      "learning_rate": 3.029183303995252e-06,
+      "loss": 0.52617874,
+      "memory(GiB)": 34.67,
+      "step": 7175,
+      "train_speed(iter/s)": 1.036483
+    },
+    {
+      "acc": 0.91279545,
+      "epoch": 12.957619477006311,
+      "grad_norm": 2.087216576180259,
+      "learning_rate": 3.02232920770612e-06,
+      "loss": 0.45612497,
+      "memory(GiB)": 34.67,
+      "step": 7180,
+      "train_speed(iter/s)": 1.03282
+    },
+    {
+      "acc": 0.91408367,
+      "epoch": 12.96663660955816,
+      "grad_norm": 2.327322093960309,
+      "learning_rate": 3.0154795167295457e-06,
+      "loss": 0.48596373,
+      "memory(GiB)": 34.67,
+      "step": 7185,
+      "train_speed(iter/s)": 1.029196
+    },
+    {
+      "acc": 0.91417637,
+      "epoch": 12.975653742110008,
+      "grad_norm": 2.8950991414330547,
+      "learning_rate": 3.0086342463195396e-06,
+      "loss": 0.46941252,
+      "memory(GiB)": 34.67,
+      "step": 7190,
+      "train_speed(iter/s)": 1.025598
+    },
+    {
+      "acc": 0.91168137,
+      "epoch": 12.984670874661857,
+      "grad_norm": 2.4907450010736816,
+      "learning_rate": 3.0017934117202598e-06,
+      "loss": 0.48299589,
+      "memory(GiB)": 34.67,
+      "step": 7195,
+      "train_speed(iter/s)": 1.02203
+    },
+    {
+      "acc": 0.91001816,
+      "epoch": 12.993688007213706,
+      "grad_norm": 3.0457935298445222,
+      "learning_rate": 2.994957028165992e-06,
+      "loss": 0.44837413,
+      "memory(GiB)": 34.67,
+      "step": 7200,
+      "train_speed(iter/s)": 1.018503
+    },
+    {
+      "epoch": 13.0,
+      "eval_acc": 0.9025353283458022,
+      "eval_loss": 0.25556719303131104,
+      "eval_runtime": 49.9473,
+      "eval_samples_per_second": 9.35,
+      "eval_steps_per_second": 1.181,
+      "step": 7204
+    },
+    {
+      "acc": 0.83156776,
+      "epoch": 13.001803426510369,
+      "grad_norm": 1.910434044119821,
+      "learning_rate": 2.9894911363060965e-06,
+      "loss": 0.37572508,
+      "memory(GiB)": 34.67,
+      "step": 7205,
+      "train_speed(iter/s)": 1.001063
+    },
+    {
+      "acc": 0.91135616,
+      "epoch": 13.010820559062218,
+      "grad_norm": 2.9010833977222426,
+      "learning_rate": 2.9826628029915806e-06,
+      "loss": 0.45042601,
+      "memory(GiB)": 34.67,
+      "step": 7210,
+      "train_speed(iter/s)": 0.997696
+    },
+    {
+      "acc": 0.91050386,
+      "epoch": 13.019837691614066,
+      "grad_norm": 2.8666784763643864,
+      "learning_rate": 2.9758389633252293e-06,
+      "loss": 0.47643499,
+      "memory(GiB)": 34.67,
+      "step": 7215,
+      "train_speed(iter/s)": 0.994355
+    },
+    {
+      "acc": 0.9150609,
+      "epoch": 13.028854824165915,
+      "grad_norm": 2.6840022848349894,
+      "learning_rate": 2.969019632503477e-06,
+      "loss": 0.44145088,
+      "memory(GiB)": 34.67,
+      "step": 7220,
+      "train_speed(iter/s)": 0.99104
+    },
+    {
+      "acc": 0.93762083,
+      "epoch": 13.037871956717764,
+      "grad_norm": 3.160010569776109,
+      "learning_rate": 2.9622048257127204e-06,
+      "loss": 0.36134422,
+      "memory(GiB)": 34.67,
+      "step": 7225,
+      "train_speed(iter/s)": 0.98775
+    },
+    {
+      "acc": 0.910851,
+      "epoch": 13.046889089269612,
+      "grad_norm": 2.27660631793261,
+      "learning_rate": 2.9553945581292793e-06,
+      "loss": 0.46951542,
+      "memory(GiB)": 34.67,
+      "step": 7230,
+      "train_speed(iter/s)": 0.984481
+    },
+    {
+      "acc": 0.90577393,
+      "epoch": 13.055906221821461,
+      "grad_norm": 2.3871135155572634,
+      "learning_rate": 2.948588844919364e-06,
+      "loss": 0.4867784,
+      "memory(GiB)": 34.67,
+      "step": 7235,
+      "train_speed(iter/s)": 0.981239
+    },
+    {
+      "acc": 0.92203779,
+      "epoch": 13.06492335437331,
+      "grad_norm": 3.545670641228415,
+      "learning_rate": 2.941787701239047e-06,
+      "loss": 0.41522141,
+      "memory(GiB)": 34.67,
+      "step": 7240,
+      "train_speed(iter/s)": 0.978024
+    },
+    {
+      "acc": 0.91149569,
+      "epoch": 13.073940486925158,
+      "grad_norm": 2.367381737816561,
+      "learning_rate": 2.93499114223422e-06,
+      "loss": 0.46987023,
+      "memory(GiB)": 34.67,
+      "step": 7245,
+      "train_speed(iter/s)": 0.974823
+    },
+    {
+      "acc": 0.93819427,
+      "epoch": 13.082957619477007,
+      "grad_norm": 1.7155186199853616,
+      "learning_rate": 2.9281991830405672e-06,
+      "loss": 0.35313714,
+      "memory(GiB)": 34.67,
+      "step": 7250,
+      "train_speed(iter/s)": 0.971652
+    },
+    {
+      "acc": 0.91345415,
+      "epoch": 13.091974752028856,
+      "grad_norm": 3.006464523365056,
+      "learning_rate": 2.921411838783525e-06,
+      "loss": 0.45448046,
+      "memory(GiB)": 34.67,
+      "step": 7255,
+      "train_speed(iter/s)": 0.968508
+    },
+    {
+      "acc": 0.90963287,
+      "epoch": 13.100991884580703,
+      "grad_norm": 3.396531814582047,
+      "learning_rate": 2.91462912457826e-06,
+      "loss": 0.44586344,
+      "memory(GiB)": 34.67,
+      "step": 7260,
+      "train_speed(iter/s)": 0.965395
+    },
+    {
+      "acc": 0.90684223,
+      "epoch": 13.110009017132551,
+      "grad_norm": 4.290854237947542,
+      "learning_rate": 2.907851055529622e-06,
+      "loss": 0.48749828,
+      "memory(GiB)": 34.67,
+      "step": 7265,
+      "train_speed(iter/s)": 0.962299
+    },
+    {
+      "acc": 0.93061886,
+      "epoch": 13.1190261496844,
+      "grad_norm": 2.0506955335973585,
+      "learning_rate": 2.90107764673212e-06,
+      "loss": 0.38269024,
+      "memory(GiB)": 34.67,
+      "step": 7270,
+      "train_speed(iter/s)": 0.959232
+    },
+    {
+      "acc": 0.91307583,
+      "epoch": 13.128043282236249,
+      "grad_norm": 2.6600976152068845,
+      "learning_rate": 2.8943089132698787e-06,
+      "loss": 0.43937092,
+      "memory(GiB)": 34.67,
+      "step": 7275,
+      "train_speed(iter/s)": 0.956189
+    },
+    {
+      "acc": 0.90629559,
+      "epoch": 13.137060414788097,
+      "grad_norm": 4.0580275056400605,
+      "learning_rate": 2.8875448702166175e-06,
+      "loss": 0.48432384,
+      "memory(GiB)": 34.67,
+      "step": 7280,
+      "train_speed(iter/s)": 0.953163
+    },
+    {
+      "acc": 0.90866852,
+      "epoch": 13.146077547339946,
+      "grad_norm": 2.7017420227692446,
+      "learning_rate": 2.8807855326356072e-06,
+      "loss": 0.4423574,
+      "memory(GiB)": 34.67,
+      "step": 7285,
+      "train_speed(iter/s)": 0.950164
+    },
+    {
+      "acc": 0.90365028,
+      "epoch": 13.155094679891794,
+      "grad_norm": 3.5664870496977796,
+      "learning_rate": 2.8740309155796436e-06,
+      "loss": 0.5123939,
+      "memory(GiB)": 34.67,
+      "step": 7290,
+      "train_speed(iter/s)": 0.947182
+    },
+    {
+      "acc": 0.90207653,
+      "epoch": 13.164111812443643,
+      "grad_norm": 3.79761429823838,
+      "learning_rate": 2.8672810340910006e-06,
+      "loss": 0.49108753,
+      "memory(GiB)": 34.67,
+      "step": 7295,
+      "train_speed(iter/s)": 0.944231
+    },
+    {
+      "acc": 0.92445564,
+      "epoch": 13.173128944995492,
+      "grad_norm": 3.2696712828467405,
+      "learning_rate": 2.860535903201418e-06,
+      "loss": 0.43105054,
+      "memory(GiB)": 34.67,
+      "step": 7300,
+      "train_speed(iter/s)": 0.941248
+    },
+    {
+      "acc": 0.91931086,
+      "epoch": 13.18214607754734,
+      "grad_norm": 3.8005889241331046,
+      "learning_rate": 2.8537955379320474e-06,
+      "loss": 0.44570789,
+      "memory(GiB)": 34.67,
+      "step": 7305,
+      "train_speed(iter/s)": 0.938165
+    },
+    {
+      "acc": 0.91292334,
+      "epoch": 13.19116321009919,
+      "grad_norm": 4.480845662924485,
+      "learning_rate": 2.8470599532934322e-06,
+      "loss": 0.42035322,
+      "memory(GiB)": 34.67,
+      "step": 7310,
+      "train_speed(iter/s)": 0.935114
+    },
+    {
+      "acc": 0.92798328,
+      "epoch": 13.200180342651038,
+      "grad_norm": 2.614446117175046,
+      "learning_rate": 2.84032916428547e-06,
+      "loss": 0.37440386,
+      "memory(GiB)": 34.67,
+      "step": 7315,
+      "train_speed(iter/s)": 0.932254
+    },
+    {
+      "acc": 0.91777925,
+      "epoch": 13.209197475202885,
+      "grad_norm": 2.4629655248966484,
+      "learning_rate": 2.8336031858973733e-06,
+      "loss": 0.43911977,
+      "memory(GiB)": 34.67,
+      "step": 7320,
+      "train_speed(iter/s)": 0.929411
+    },
+    {
+      "acc": 0.91801252,
+      "epoch": 13.218214607754733,
+      "grad_norm": 2.4837495393468174,
+      "learning_rate": 2.8268820331076464e-06,
+      "loss": 0.4110014,
+      "memory(GiB)": 34.67,
+      "step": 7325,
+      "train_speed(iter/s)": 0.926584
+    },
+    {
+      "acc": 0.92134676,
+      "epoch": 13.227231740306582,
+      "grad_norm": 2.4032660248792648,
+      "learning_rate": 2.8201657208840456e-06,
+      "loss": 0.42779732,
+      "memory(GiB)": 34.67,
+      "step": 7330,
+      "train_speed(iter/s)": 0.923779
+    },
+    {
+      "acc": 0.90932083,
+      "epoch": 13.23624887285843,
+      "grad_norm": 3.394816327004731,
+      "learning_rate": 2.8134542641835507e-06,
+      "loss": 0.44342723,
+      "memory(GiB)": 34.67,
+      "step": 7335,
+      "train_speed(iter/s)": 0.921001
+    },
+    {
+      "acc": 0.92459488,
+      "epoch": 13.24526600541028,
+      "grad_norm": 2.1125145267494436,
+      "learning_rate": 2.8067476779523206e-06,
+      "loss": 0.40396466,
+      "memory(GiB)": 34.67,
+      "step": 7340,
+      "train_speed(iter/s)": 0.918242
+    },
+    {
+      "acc": 0.92494678,
+      "epoch": 13.254283137962128,
+      "grad_norm": 2.646012026414015,
+      "learning_rate": 2.8000459771256757e-06,
+      "loss": 0.37966986,
+      "memory(GiB)": 34.67,
+      "step": 7345,
+      "train_speed(iter/s)": 0.915503
+    },
+    {
+      "acc": 0.91678486,
+      "epoch": 13.263300270513977,
+      "grad_norm": 2.642400969521799,
+      "learning_rate": 2.793349176628051e-06,
+      "loss": 0.41771069,
+      "memory(GiB)": 34.67,
+      "step": 7350,
+      "train_speed(iter/s)": 0.912783
+    },
+    {
+      "acc": 0.91658487,
+      "epoch": 13.272317403065825,
+      "grad_norm": 2.281304435325576,
+      "learning_rate": 2.7866572913729763e-06,
+      "loss": 0.46291165,
+      "memory(GiB)": 34.67,
+      "step": 7355,
+      "train_speed(iter/s)": 0.910079
+    },
+    {
+      "acc": 0.91275826,
+      "epoch": 13.281334535617674,
+      "grad_norm": 2.70930507000483,
+      "learning_rate": 2.7799703362630258e-06,
+      "loss": 0.458354,
+      "memory(GiB)": 34.67,
+      "step": 7360,
+      "train_speed(iter/s)": 0.907393
+    },
+    {
+      "acc": 0.91499748,
+      "epoch": 13.290351668169523,
+      "grad_norm": 3.7940222227937754,
+      "learning_rate": 2.773288326189798e-06,
+      "loss": 0.47010036,
+      "memory(GiB)": 34.67,
+      "step": 7365,
+      "train_speed(iter/s)": 0.904731
+    },
+    {
+      "acc": 0.92498779,
+      "epoch": 13.299368800721371,
+      "grad_norm": 2.982398889619566,
+      "learning_rate": 2.7666112760338853e-06,
+      "loss": 0.43736539,
+      "memory(GiB)": 34.67,
+      "step": 7370,
+      "train_speed(iter/s)": 0.902095
+    },
+    {
+      "acc": 0.91043816,
+      "epoch": 13.30838593327322,
+      "grad_norm": 2.0679015274794486,
+      "learning_rate": 2.7599392006648285e-06,
+      "loss": 0.46530938,
+      "memory(GiB)": 34.67,
+      "step": 7375,
+      "train_speed(iter/s)": 0.899467
+    },
+    {
+      "acc": 0.92218437,
+      "epoch": 13.317403065825067,
+      "grad_norm": 1.8848611174016012,
+      "learning_rate": 2.753272114941087e-06,
+      "loss": 0.4195118,
+      "memory(GiB)": 34.67,
+      "step": 7380,
+      "train_speed(iter/s)": 0.896868
+    },
+    {
+      "acc": 0.93011036,
+      "epoch": 13.326420198376915,
+      "grad_norm": 2.7178073478219082,
+      "learning_rate": 2.746610033710013e-06,
+      "loss": 0.36692064,
+      "memory(GiB)": 34.67,
+      "step": 7385,
+      "train_speed(iter/s)": 0.894284
+    },
+    {
+      "acc": 0.90709066,
+      "epoch": 13.335437330928764,
+      "grad_norm": 2.56576319099396,
+      "learning_rate": 2.7399529718078143e-06,
+      "loss": 0.50453234,
+      "memory(GiB)": 34.67,
+      "step": 7390,
+      "train_speed(iter/s)": 0.891719
+    },
+    {
+      "acc": 0.90788584,
+      "epoch": 13.344454463480613,
+      "grad_norm": 2.340105644715265,
+      "learning_rate": 2.7333009440595216e-06,
+      "loss": 0.50700455,
+      "memory(GiB)": 34.67,
+      "step": 7395,
+      "train_speed(iter/s)": 0.889171
+    },
+    {
+      "acc": 0.90255871,
+      "epoch": 13.353471596032461,
+      "grad_norm": 2.7141721941934924,
+      "learning_rate": 2.726653965278947e-06,
+      "loss": 0.46706529,
+      "memory(GiB)": 34.67,
+      "step": 7400,
+      "train_speed(iter/s)": 0.886633
+    },
+    {
+      "acc": 0.92647762,
+      "epoch": 13.36248872858431,
+      "grad_norm": 4.133163480197946,
+      "learning_rate": 2.720012050268667e-06,
+      "loss": 0.39142969,
+      "memory(GiB)": 34.67,
+      "step": 7405,
+      "train_speed(iter/s)": 0.884115
+    },
+    {
+      "acc": 0.91706448,
+      "epoch": 13.371505861136159,
+      "grad_norm": 2.8733374569819206,
+      "learning_rate": 2.713375213819978e-06,
+      "loss": 0.4484066,
+      "memory(GiB)": 34.67,
+      "step": 7410,
+      "train_speed(iter/s)": 0.881621
+    },
+    {
+      "acc": 0.89831924,
+      "epoch": 13.380522993688007,
+      "grad_norm": 3.7669824367312152,
+      "learning_rate": 2.7067434707128708e-06,
+      "loss": 0.49606962,
+      "memory(GiB)": 34.67,
+      "step": 7415,
+      "train_speed(iter/s)": 0.879144
+    },
+    {
+      "acc": 0.89532213,
+      "epoch": 13.389540126239856,
+      "grad_norm": 3.3646782634934875,
+      "learning_rate": 2.7001168357159846e-06,
+      "loss": 0.49138098,
+      "memory(GiB)": 34.67,
+      "step": 7420,
+      "train_speed(iter/s)": 0.876686
+    },
+    {
+      "acc": 0.92779675,
+      "epoch": 13.398557258791705,
+      "grad_norm": 2.9373135278479396,
+      "learning_rate": 2.69349532358659e-06,
+      "loss": 0.38355405,
+      "memory(GiB)": 34.67,
+      "step": 7425,
+      "train_speed(iter/s)": 0.874246
+    },
+    {
+      "acc": 0.92377272,
+      "epoch": 13.407574391343553,
+      "grad_norm": 2.9765340958395754,
+      "learning_rate": 2.686878949070549e-06,
+      "loss": 0.41119652,
+      "memory(GiB)": 34.67,
+      "step": 7430,
+      "train_speed(iter/s)": 0.87182
+    },
+    {
+      "acc": 0.91067591,
+      "epoch": 13.416591523895402,
+      "grad_norm": 3.705403956327573,
+      "learning_rate": 2.6802677269022815e-06,
+      "loss": 0.45417252,
+      "memory(GiB)": 34.67,
+      "step": 7435,
+      "train_speed(iter/s)": 0.869415
+    },
+    {
+      "acc": 0.92136078,
+      "epoch": 13.425608656447249,
+      "grad_norm": 3.2789817249364375,
+      "learning_rate": 2.6736616718047286e-06,
+      "loss": 0.40200434,
+      "memory(GiB)": 34.67,
+      "step": 7440,
+      "train_speed(iter/s)": 0.867017
+    },
+    {
+      "acc": 0.91240864,
+      "epoch": 13.434625788999098,
+      "grad_norm": 2.2247812089028596,
+      "learning_rate": 2.6670607984893315e-06,
+      "loss": 0.46096272,
+      "memory(GiB)": 34.67,
+      "step": 7445,
+      "train_speed(iter/s)": 0.864643
+    },
+    {
+      "acc": 0.90765324,
+      "epoch": 13.443642921550946,
+      "grad_norm": 2.5340394179146974,
+      "learning_rate": 2.6604651216559887e-06,
+      "loss": 0.41848903,
+      "memory(GiB)": 34.67,
+      "step": 7450,
+      "train_speed(iter/s)": 0.862279
+    },
+    {
+      "acc": 0.90695438,
+      "epoch": 13.452660054102795,
+      "grad_norm": 2.563710193611515,
+      "learning_rate": 2.653874655993026e-06,
+      "loss": 0.49381599,
+      "memory(GiB)": 34.67,
+      "step": 7455,
+      "train_speed(iter/s)": 0.859933
+    },
+    {
+      "acc": 0.91303482,
+      "epoch": 13.461677186654644,
+      "grad_norm": 3.047339345762434,
+      "learning_rate": 2.647289416177167e-06,
+      "loss": 0.44869227,
+      "memory(GiB)": 34.67,
+      "step": 7460,
+      "train_speed(iter/s)": 0.857597
+    },
+    {
+      "acc": 0.92487192,
+      "epoch": 13.470694319206492,
+      "grad_norm": 2.1385095794722724,
+      "learning_rate": 2.64070941687349e-06,
+      "loss": 0.40038085,
+      "memory(GiB)": 34.67,
+      "step": 7465,
+      "train_speed(iter/s)": 0.855277
+    },
+    {
+      "acc": 0.90575371,
+      "epoch": 13.479711451758341,
+      "grad_norm": 3.0483696799435225,
+      "learning_rate": 2.6341346727354113e-06,
+      "loss": 0.50467086,
+      "memory(GiB)": 34.67,
+      "step": 7470,
+      "train_speed(iter/s)": 0.852982
+    },
+    {
+      "acc": 0.91957855,
+      "epoch": 13.48872858431019,
+      "grad_norm": 3.6440923105623684,
+      "learning_rate": 2.62756519840464e-06,
+      "loss": 0.44822865,
+      "memory(GiB)": 34.67,
+      "step": 7475,
+      "train_speed(iter/s)": 0.850704
+    },
+    {
+      "acc": 0.92911835,
+      "epoch": 13.497745716862038,
+      "grad_norm": 3.2803904544089213,
+      "learning_rate": 2.6210010085111507e-06,
+      "loss": 0.37346206,
+      "memory(GiB)": 34.67,
+      "step": 7480,
+      "train_speed(iter/s)": 0.848441
+    },
+    {
+      "acc": 0.91992598,
+      "epoch": 13.506762849413887,
+      "grad_norm": 2.254405202090834,
+      "learning_rate": 2.6144421176731503e-06,
+      "loss": 0.4058012,
+      "memory(GiB)": 34.67,
+      "step": 7485,
+      "train_speed(iter/s)": 0.846191
+    },
+    {
+      "acc": 0.92690392,
+      "epoch": 13.515779981965736,
+      "grad_norm": 2.8697510836962388,
+      "learning_rate": 2.6078885404970415e-06,
+      "loss": 0.40351672,
+      "memory(GiB)": 34.67,
+      "step": 7490,
+      "train_speed(iter/s)": 0.84395
+    },
+    {
+      "acc": 0.91610775,
+      "epoch": 13.524797114517582,
+      "grad_norm": 2.051569997394548,
+      "learning_rate": 2.6013402915774017e-06,
+      "loss": 0.46603556,
+      "memory(GiB)": 34.67,
+      "step": 7495,
+      "train_speed(iter/s)": 0.84172
+    },
+    {
+      "acc": 0.91617641,
+      "epoch": 13.533814247069431,
+      "grad_norm": 2.286165118967432,
+      "learning_rate": 2.59479738549693e-06,
+      "loss": 0.45086269,
+      "memory(GiB)": 34.67,
+      "step": 7500,
+      "train_speed(iter/s)": 0.839511
+    },
+    {
+      "acc": 0.93104916,
+      "epoch": 13.54283137962128,
+      "grad_norm": 3.8643444719875646,
+      "learning_rate": 2.588259836826437e-06,
+      "loss": 0.37835269,
+      "memory(GiB)": 34.67,
+      "step": 7505,
+      "train_speed(iter/s)": 0.837317
+    },
+    {
+      "acc": 0.92141943,
+      "epoch": 13.551848512173128,
+      "grad_norm": 2.225645199352156,
+      "learning_rate": 2.5817276601248016e-06,
+      "loss": 0.44836111,
+      "memory(GiB)": 34.67,
+      "step": 7510,
+      "train_speed(iter/s)": 0.835139
+    },
+    {
+      "acc": 0.9025341,
+      "epoch": 13.560865644724977,
+      "grad_norm": 3.4844716672940432,
+      "learning_rate": 2.5752008699389348e-06,
+      "loss": 0.50761642,
+      "memory(GiB)": 34.67,
+      "step": 7515,
+      "train_speed(iter/s)": 0.832969
+    },
+    {
+      "acc": 0.91783009,
+      "epoch": 13.569882777276826,
+      "grad_norm": 3.326678568900906,
+      "learning_rate": 2.5686794808037574e-06,
+      "loss": 0.46847334,
+      "memory(GiB)": 34.67,
+      "step": 7520,
+      "train_speed(iter/s)": 0.830819
+    },
+    {
+      "acc": 0.91352987,
+      "epoch": 13.578899909828674,
+      "grad_norm": 2.8228888062944786,
+      "learning_rate": 2.5621635072421575e-06,
+      "loss": 0.46015668,
+      "memory(GiB)": 34.67,
+      "step": 7525,
+      "train_speed(iter/s)": 0.828649
+    },
+    {
+      "acc": 0.92407179,
+      "epoch": 13.587917042380523,
+      "grad_norm": 3.006678121026572,
+      "learning_rate": 2.5556529637649653e-06,
+      "loss": 0.45808477,
+      "memory(GiB)": 34.67,
+      "step": 7530,
+      "train_speed(iter/s)": 0.826424
+    },
+    {
+      "acc": 0.92271099,
+      "epoch": 13.596934174932372,
+      "grad_norm": 2.3647315636854396,
+      "learning_rate": 2.5491478648709196e-06,
+      "loss": 0.40352392,
+      "memory(GiB)": 34.67,
+      "step": 7535,
+      "train_speed(iter/s)": 0.824268
+    },
+    {
+      "acc": 0.90452938,
+      "epoch": 13.60595130748422,
+      "grad_norm": 2.586311310712336,
+      "learning_rate": 2.5426482250466338e-06,
+      "loss": 0.53205042,
+      "memory(GiB)": 34.67,
+      "step": 7540,
+      "train_speed(iter/s)": 0.822175
+    },
+    {
+      "acc": 0.91857758,
+      "epoch": 13.614968440036069,
+      "grad_norm": 1.7748948002446534,
+      "learning_rate": 2.5361540587665605e-06,
+      "loss": 0.42117982,
+      "memory(GiB)": 34.67,
+      "step": 7545,
+      "train_speed(iter/s)": 0.820091
+    },
+    {
+      "acc": 0.92564392,
+      "epoch": 13.623985572587918,
+      "grad_norm": 2.416666583742134,
+      "learning_rate": 2.5296653804929667e-06,
+      "loss": 0.37726135,
+      "memory(GiB)": 34.67,
+      "step": 7550,
+      "train_speed(iter/s)": 0.818016
+    },
+    {
+      "acc": 0.90426903,
+      "epoch": 13.633002705139766,
+      "grad_norm": 2.4728096245909317,
+      "learning_rate": 2.5231822046758976e-06,
+      "loss": 0.51140699,
+      "memory(GiB)": 34.67,
+      "step": 7555,
+      "train_speed(iter/s)": 0.815963
+    },
+    {
+      "acc": 0.91746035,
+      "epoch": 13.642019837691613,
+      "grad_norm": 2.2490695665177975,
+      "learning_rate": 2.5167045457531447e-06,
+      "loss": 0.45246611,
+      "memory(GiB)": 34.67,
+      "step": 7560,
+      "train_speed(iter/s)": 0.813926
+    },
+    {
+      "acc": 0.92741585,
+      "epoch": 13.651036970243462,
+      "grad_norm": 3.770299354880114,
+      "learning_rate": 2.510232418150211e-06,
+      "loss": 0.37998924,
+      "memory(GiB)": 34.67,
+      "step": 7565,
+      "train_speed(iter/s)": 0.811899
+    },
+    {
+      "acc": 0.9042696,
+      "epoch": 13.66005410279531,
+      "grad_norm": 2.426793152085983,
+      "learning_rate": 2.503765836280283e-06,
+      "loss": 0.50290551,
+      "memory(GiB)": 34.67,
+      "step": 7570,
+      "train_speed(iter/s)": 0.809883
+    },
+    {
+      "acc": 0.92482624,
+      "epoch": 13.66907123534716,
+      "grad_norm": 3.5893481594395875,
+      "learning_rate": 2.4973048145441984e-06,
+      "loss": 0.38070757,
+      "memory(GiB)": 34.67,
+      "step": 7575,
+      "train_speed(iter/s)": 0.807877
+    },
+    {
+      "acc": 0.90683136,
+      "epoch": 13.678088367899008,
+      "grad_norm": 2.4112487872224104,
+      "learning_rate": 2.490849367330412e-06,
+      "loss": 0.45160856,
+      "memory(GiB)": 34.67,
+      "step": 7580,
+      "train_speed(iter/s)": 0.805888
+    },
+    {
+      "acc": 0.91481924,
+      "epoch": 13.687105500450857,
+      "grad_norm": 3.9976426199831074,
+      "learning_rate": 2.484399509014962e-06,
+      "loss": 0.44831839,
+      "memory(GiB)": 34.67,
+      "step": 7585,
+      "train_speed(iter/s)": 0.803907
+    },
+    {
+      "acc": 0.91878014,
+      "epoch": 13.696122633002705,
+      "grad_norm": 2.028143131118245,
+      "learning_rate": 2.4779552539614447e-06,
+      "loss": 0.41924124,
+      "memory(GiB)": 34.67,
+      "step": 7590,
+      "train_speed(iter/s)": 0.801939
+    },
+    {
+      "acc": 0.91641684,
+      "epoch": 13.705139765554554,
+      "grad_norm": 3.8461718844426542,
+      "learning_rate": 2.4715166165209736e-06,
+      "loss": 0.42823806,
+      "memory(GiB)": 34.67,
+      "step": 7595,
+      "train_speed(iter/s)": 0.799981
+    },
+    {
+      "acc": 0.91030054,
+      "epoch": 13.714156898106403,
+      "grad_norm": 2.209311395879964,
+      "learning_rate": 2.465083611032158e-06,
+      "loss": 0.47398629,
+      "memory(GiB)": 34.67,
+      "step": 7600,
+      "train_speed(iter/s)": 0.798037
+    },
+    {
+      "acc": 0.93113747,
+      "epoch": 13.723174030658251,
+      "grad_norm": 2.5553222520456518,
+      "learning_rate": 2.4586562518210554e-06,
+      "loss": 0.39157214,
+      "memory(GiB)": 34.67,
+      "step": 7605,
+      "train_speed(iter/s)": 0.79611
+    },
+    {
+      "acc": 0.91998997,
+      "epoch": 13.7321911632101,
+      "grad_norm": 2.7813427865917446,
+      "learning_rate": 2.4522345532011583e-06,
+      "loss": 0.42484126,
+      "memory(GiB)": 34.67,
+      "step": 7610,
+      "train_speed(iter/s)": 0.794195
+    },
+    {
+      "acc": 0.91799946,
+      "epoch": 13.741208295761947,
+      "grad_norm": 2.565290365377133,
+      "learning_rate": 2.445818529473349e-06,
+      "loss": 0.44959812,
+      "memory(GiB)": 34.67,
+      "step": 7615,
+      "train_speed(iter/s)": 0.792289
+    },
+    {
+      "acc": 0.93449316,
+      "epoch": 13.750225428313795,
+      "grad_norm": 2.7806118998990756,
+      "learning_rate": 2.4394081949258754e-06,
+      "loss": 0.36552901,
+      "memory(GiB)": 34.67,
+      "step": 7620,
+      "train_speed(iter/s)": 0.790394
+    },
+    {
+      "acc": 0.92006874,
+      "epoch": 13.759242560865644,
+      "grad_norm": 2.6335169545111685,
+      "learning_rate": 2.43300356383431e-06,
+      "loss": 0.42915936,
+      "memory(GiB)": 34.67,
+      "step": 7625,
+      "train_speed(iter/s)": 0.788513
+    },
+    {
+      "acc": 0.91715298,
+      "epoch": 13.768259693417493,
+      "grad_norm": 2.6110020416685806,
+      "learning_rate": 2.4266046504615283e-06,
+      "loss": 0.43826027,
+      "memory(GiB)": 34.67,
+      "step": 7630,
+      "train_speed(iter/s)": 0.786641
+    },
+    {
+      "acc": 0.9005537,
+      "epoch": 13.777276825969341,
+      "grad_norm": 2.389448526677138,
+      "learning_rate": 2.420211469057672e-06,
+      "loss": 0.49912429,
+      "memory(GiB)": 34.67,
+      "step": 7635,
+      "train_speed(iter/s)": 0.784778
+    },
+    {
+      "acc": 0.93391418,
+      "epoch": 13.78629395852119,
+      "grad_norm": 2.3303458261589105,
+      "learning_rate": 2.4138240338601187e-06,
+      "loss": 0.39306188,
+      "memory(GiB)": 34.67,
+      "step": 7640,
+      "train_speed(iter/s)": 0.782927
+    },
+    {
+      "acc": 0.91246166,
+      "epoch": 13.795311091073039,
+      "grad_norm": 3.345441834133012,
+      "learning_rate": 2.407442359093448e-06,
+      "loss": 0.4463088,
+      "memory(GiB)": 34.67,
+      "step": 7645,
+      "train_speed(iter/s)": 0.781091
+    },
+    {
+      "acc": 0.92767334,
+      "epoch": 13.804328223624887,
+      "grad_norm": 2.311986915142374,
+      "learning_rate": 2.40106645896941e-06,
+      "loss": 0.40542097,
+      "memory(GiB)": 34.67,
+      "step": 7650,
+      "train_speed(iter/s)": 0.779264
+    },
+    {
+      "acc": 0.92072134,
+      "epoch": 13.813345356176736,
+      "grad_norm": 2.721893786946378,
+      "learning_rate": 2.394696347686897e-06,
+      "loss": 0.4374537,
+      "memory(GiB)": 34.67,
+      "step": 7655,
+      "train_speed(iter/s)": 0.777446
+    },
+    {
+      "acc": 0.89793825,
+      "epoch": 13.822362488728585,
+      "grad_norm": 3.04927251622692,
+      "learning_rate": 2.388332039431909e-06,
+      "loss": 0.51242447,
+      "memory(GiB)": 34.67,
+      "step": 7660,
+      "train_speed(iter/s)": 0.775638
+    },
+    {
+      "acc": 0.91029463,
+      "epoch": 13.831379621280433,
+      "grad_norm": 2.330530114774836,
+      "learning_rate": 2.381973548377526e-06,
+      "loss": 0.4661911,
+      "memory(GiB)": 34.67,
+      "step": 7665,
+      "train_speed(iter/s)": 0.773837
+    },
+    {
+      "acc": 0.91078424,
+      "epoch": 13.840396753832282,
+      "grad_norm": 3.0249717723702103,
+      "learning_rate": 2.3756208886838657e-06,
+      "loss": 0.45671978,
+      "memory(GiB)": 34.67,
+      "step": 7670,
+      "train_speed(iter/s)": 0.772046
+    },
+    {
+      "acc": 0.92062321,
+      "epoch": 13.84941388638413,
+      "grad_norm": 2.3289482832980455,
+      "learning_rate": 2.3692740744980665e-06,
+      "loss": 0.45508952,
+      "memory(GiB)": 34.67,
+      "step": 7675,
+      "train_speed(iter/s)": 0.770273
+    },
+    {
+      "acc": 0.9151104,
+      "epoch": 13.858431018935978,
+      "grad_norm": 1.991707358142325,
+      "learning_rate": 2.3629331199542472e-06,
+      "loss": 0.45227852,
+      "memory(GiB)": 34.67,
+      "step": 7680,
+      "train_speed(iter/s)": 0.768512
+    },
+    {
+      "acc": 0.92223511,
+      "epoch": 13.867448151487826,
+      "grad_norm": 2.8714618075645575,
+      "learning_rate": 2.3565980391734785e-06,
+      "loss": 0.41313591,
+      "memory(GiB)": 34.67,
+      "step": 7685,
+      "train_speed(iter/s)": 0.766755
+    },
+    {
+      "acc": 0.91029892,
+      "epoch": 13.876465284039675,
+      "grad_norm": 3.8192779037514044,
+      "learning_rate": 2.350268846263745e-06,
+      "loss": 0.47502518,
+      "memory(GiB)": 34.67,
+      "step": 7690,
+      "train_speed(iter/s)": 0.765014
+    },
+    {
+      "acc": 0.90931625,
+      "epoch": 13.885482416591524,
+      "grad_norm": 3.6311347936367198,
+      "learning_rate": 2.3439455553199234e-06,
+      "loss": 0.52817411,
+      "memory(GiB)": 34.67,
+      "step": 7695,
+      "train_speed(iter/s)": 0.763279
+    },
+    {
+      "acc": 0.90909996,
+      "epoch": 13.894499549143372,
+      "grad_norm": 3.070435648155786,
+      "learning_rate": 2.337628180423752e-06,
+      "loss": 0.47671447,
+      "memory(GiB)": 34.67,
+      "step": 7700,
+      "train_speed(iter/s)": 0.76155
+    },
+    {
+      "acc": 0.91981907,
+      "epoch": 13.903516681695221,
+      "grad_norm": 2.9136339190019367,
+      "learning_rate": 2.3313167356437847e-06,
+      "loss": 0.4204464,
+      "memory(GiB)": 34.67,
+      "step": 7705,
+      "train_speed(iter/s)": 0.759827
+    },
+    {
+      "acc": 0.9279768,
+      "epoch": 13.91253381424707,
+      "grad_norm": 1.7584487100424862,
+      "learning_rate": 2.3250112350353738e-06,
+      "loss": 0.3965879,
+      "memory(GiB)": 34.67,
+      "step": 7710,
+      "train_speed(iter/s)": 0.758123
+    },
+    {
+      "acc": 0.92543163,
+      "epoch": 13.921550946798918,
+      "grad_norm": 2.074743460894259,
+      "learning_rate": 2.318711692640635e-06,
+      "loss": 0.40781312,
+      "memory(GiB)": 34.67,
+      "step": 7715,
+      "train_speed(iter/s)": 0.756432
+    },
+    {
+      "acc": 0.92013721,
+      "epoch": 13.930568079350767,
+      "grad_norm": 2.391497930161413,
+      "learning_rate": 2.312418122488414e-06,
+      "loss": 0.40785055,
+      "memory(GiB)": 34.67,
+      "step": 7720,
+      "train_speed(iter/s)": 0.754747
+    },
+    {
+      "acc": 0.90895977,
+      "epoch": 13.939585211902616,
+      "grad_norm": 2.282832199655629,
+      "learning_rate": 2.306130538594259e-06,
+      "loss": 0.48620806,
+      "memory(GiB)": 34.67,
+      "step": 7725,
+      "train_speed(iter/s)": 0.75307
+    },
+    {
+      "acc": 0.92811804,
+      "epoch": 13.948602344454464,
+      "grad_norm": 3.2542288268715436,
+      "learning_rate": 2.2998489549603794e-06,
+      "loss": 0.36995916,
+      "memory(GiB)": 34.67,
+      "step": 7730,
+      "train_speed(iter/s)": 0.751402
+    },
+    {
+      "acc": 0.92376108,
+      "epoch": 13.957619477006311,
+      "grad_norm": 1.9798170143802984,
+      "learning_rate": 2.2935733855756316e-06,
+      "loss": 0.42149258,
+      "memory(GiB)": 34.67,
+      "step": 7735,
+      "train_speed(iter/s)": 0.749747
+    },
+    {
+      "acc": 0.92298889,
+      "epoch": 13.96663660955816,
+      "grad_norm": 2.7042116336273865,
+      "learning_rate": 2.287303844415473e-06,
+      "loss": 0.40348473,
+      "memory(GiB)": 34.67,
+      "step": 7740,
+      "train_speed(iter/s)": 0.748106
+    },
+    {
+      "acc": 0.91418896,
+      "epoch": 13.975653742110008,
+      "grad_norm": 2.2087753277719946,
+      "learning_rate": 2.281040345441941e-06,
+      "loss": 0.4406374,
+      "memory(GiB)": 34.67,
+      "step": 7745,
+      "train_speed(iter/s)": 0.746471
+    },
+    {
+      "acc": 0.90041313,
+      "epoch": 13.984670874661857,
+      "grad_norm": 3.272289949082435,
+      "learning_rate": 2.274782902603609e-06,
+      "loss": 0.51967297,
+      "memory(GiB)": 34.67,
+      "step": 7750,
+      "train_speed(iter/s)": 0.744826
+    },
+    {
+      "acc": 0.91371441,
+      "epoch": 13.993688007213706,
+      "grad_norm": 3.8788683819560346,
+      "learning_rate": 2.2685315298355725e-06,
+      "loss": 0.48910103,
+      "memory(GiB)": 34.67,
+      "step": 7755,
+      "train_speed(iter/s)": 0.743147
+    },
+    {
+      "epoch": 14.0,
+      "eval_acc": 0.9115752285951787,
+      "eval_loss": 0.24021737277507782,
+      "eval_runtime": 49.8939,
+      "eval_samples_per_second": 9.36,
+      "eval_steps_per_second": 1.183,
+      "step": 7759
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 11080,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 20,
+  "save_steps": 200,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 5105557816049664.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}